BACKLOG P: AD sign-in to Slurm user, tried end to end

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

.gitea/workflows/ci.yml

@@ -34,7 +34,11 @@ jobs:
         run: go vet ./...
 
       - name: Test with the race detector
-        run: go test -race ./...
+        # Under the job's limit, so a package that hangs panics with the
+        # test it was in and where every goroutine waited, rather than the
+        # job ending at its deadline with nothing said: from 2026-10-15 every
+        # run did, printing nothing after internal/release.
+        run: go test -race -timeout 12m ./...
 
       - name: Build
         run: go build ./...

.gitea/workflows/engine-releases.yml

@@ -0,0 +1,20 @@
+# Fails when an engine llmgw follows release by release (FreeToken) has
+# published a version newer than the one DefaultEngines installs: the failed
+# run is the notice to check the new version and move to it. See
+# scripts/engine_releases.py.
+name: Engine releases
+
+on:
+  schedule:
+    # Mondays 06:00 UTC.
+    - cron: "0 6 * * 1"
+  workflow_dispatch:
+
+jobs:
+  check:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Compare with PyPI
+        run: python3 scripts/engine_releases.py

.gitignore

@@ -9,3 +9,5 @@ coverage.out
 /terraform-provider-llmgw/terraform-provider-llmgw
 /terraform-provider-llmgw/examples/.terraform*
 /terraform-provider-llmgw/examples/terraform.tfstate*
+# What a development server run from the checkout writes beside it.
+/secrets/

BACKLOG.md

@@ -15,6 +15,66 @@ Conventions this codebase keeps, which every item here must keep too:
 - `go test ./...`, `go vet ./...` and `govulncheck ./...` pass. No release is
   cut and nothing is deployed to the PoC cluster without being told to.
 
+## Open, in order (2026-10-20)
+
+What is left, most needed first. The detail is in the item each names.
+
+**Now: what goes wrong where it is installed.**
+1. ~~**J8**~~: done 2026-10-20. `ctl guard judge check`, and init times the judge. Left of **J3**: its load test
+   against judges on the control plane, which was stopped half way. Judges on a 4-core node of their own refused
+   0/60 at 2 at once, 13/60 at 4 and 42/60 at 10.
+2. ~~**I2 P2**~~: done 2026-10-20: the guardrails both ways, a rate per caller and server, a breaker. A portal page
+   for the servers is left, with P3 (stdio, per-person OAuth).
+3. **K**: fixes from the fabric test the operator runs with docs/large-models.md.
+
+**Soon: the customer's experience, and HA.**
+4. ~~**C24**~~: done 2026-10-21. The store's, catalog's and mirror's errors, percent signs, footer times and stale
+   forms are in the page's language. Left: the /docs documents, and a language kept per person (cookie only today).
+5. ~~**J4**~~: done 2026-10-29. The done-criterion held on the PoC: with mirror 146 stopped, `ctl model serve` of a
+   new model was fetched by the second mirror (VM 167, 10.0.0.152), placed and answered; a node without a file
+   146's address names was given 152's; what was added meanwhile reached 146 when it came back; a removal was
+   carried both ways. Found and fixed there: add-peer on a mirror machine of its own, files without checksums never
+   copied, and `mirror remove` neither recorded nor seen by the running mirror.
+6. **F5**: large imports. Done 2026-10-29: the gateway embeds what `llmgw-kb` sends as text; the slots and
+   halfvec were done before. Left: a larger store machine before a million chunks (the operator's), and the
+   embedding model on a GPU for a real import.
+
+**From Kong** (section L): L1–L5 and L9 done (v0.15.0); left of L4 is routing by cost between outside
+providers, of L5 cache-write and audio prices. L6 and L7 are llmgw-80's; L8 (A2A) when asked.
+
+**Models we can trust** (section M, 2026-10-25): M1 done (pickles refused; the explorer lists only what this
+cluster's hardware holds, with CPU and GPU tok/s). Next M2 (every download pinned and checked against Hugging
+Face, done 2026-10-30), then M5 (engines without the internet), M3, M4, M6.
+
+**Scale.** O: more than one control plane answering clients; O1 (measuring one router's ceiling) first.
+
+**Governance.**
+7. ~~**C1**~~: done 2026-11-08. An organization's admin (`--role admin` on an org membership) changes its teams,
+   projects, people and keys from /org; viewers are drawn no forms; forms made compact.
+8. **Audit before and after**: what a change changed, and changes made with the CLI recorded too.
+
+**When there is room.**
+9. ~~**I1**~~: done 2026-10-25 (in memory, not pgvector); left: the 40-rewordings run on the PoC.
+10. **H1**: tried on the PoC.
+11. **C4**: Parquet, and a scheduled push for usage export.
+
+**When needed.**
+12. **H5**: licences in video and audio. 13. **H3**: replicas shared, the catalog side. 14. **J7**: nginx in front
+of several control planes. 15. **J6**: rate limits per process (a note).
+
+**Done since 2026-10-16:**
+- J1, J5, I3, I4, J2.
+- Active Directory and Entra ID, tried against a Samba DC, with the console's Sign-in screen.
+- J3's code.
+- K's code and docs/large-models.md.
+- I2 P1 and the cluster's own read-only MCP server (llmgw:admin).
+- The node console's Overview; a node's page showing its version and the models on each GPU.
+
+**Not code, the operator's to decide or do:**
+- Where Alertmanager sends (its receiver is "none"): with a judge down, nobody hears.
+- The fabric test for K.
+- The PoC runs a dev build of main (v0.15.0-7-ga583359) on its control planes and nodes, the mirror v0.15.0.
+
 ---
 
 ## A. Security — do first
@@ -376,9 +436,9 @@ portal is not done — the page shows; the console and ctl change.
    API middleware and the /admin pages (with each change's result), `ctl
    audit`, the console's Audit view; staff roles viewer / operator /
    billing / admin. Left: before/after values (the request's fields are
-   recorded, not the state they replaced); a tenant-admin role that changes
-   its own organization (managers still only view it); hiding the forms a
-   viewer cannot submit. **Audit log and roles (M).** Today "audit" is slog lines
+   recorded, not the state they replaced). Done 2026-11-08: an organization's
+   admin who changes its own organization from /org (managers still only view
+   it), and the forms a viewer cannot submit no longer drawn. **Audit log and roles (M).** Today "audit" is slog lines
    (`internal/web/admin_users.go:403-480`) and roles are the `IsAdmin` flag
    plus an org "manager" who can view. Add `audit_events` (who, what,
    before/after, source) written by the `adminAction`/`tenancyAction`
@@ -403,8 +463,9 @@ portal is not done — the page shows; the console and ctl change.
 5. **Done (branch `features-c`):** `notify.webhooks` (model parked, node
    lost, budget spent, update available; signed JSON or Slack; retries,
    hourly de-duplication), `ctl notify test`, Alertmanager receiver
-   templates. Left: `server init` and the console do not yet ask for a
-   receiver — the templates are filled in by hand. Events are claimed in
+   templates. `server init` and both setup screens ask for webhooks and test
+   each (d1e9186, 2026-10-14); with none set, `ctl status` and the console's
+   Overview say the alerts go nowhere (2026-11-12). Events are claimed in
    the store before they are sent (2026-10-10), so two control planes send
    each once.
    **Alerting out of the box (S).** `deploy/monitoring/alertmanager.yml`
@@ -1073,13 +1134,19 @@ four texts per pass, while llmgw-kb sends 32 a request from 6 workers. The
 usage rows show 4.6 s a request on average, 1.8 s of it in the gateway's
 queue and 0.04 s in the guard. The GPU sat at 18–50% and the store machine
 at a third of its CPUs.
-- An embedding model's profile with many slots by default (`--parallel`
+- Done for `ctl model serve` (32 slots of 1024 tokens; `ctl model context
+  MODEL 1024 --slots 32` resizes one served before): an embedding model's profile with many slots by default (`--parallel`
   32–64, `--ctx-size` to match, `--batch-size`/`--ubatch-size` at least the
   longest chunk), or `ctl model serve` choosing that for a model with
   `capabilities.embeddings`. With the default 512, one 516-token chunk failed.
-- Embedding on the gateway's side of a document PUT, so the client sends
-  text only: today an embeddings answer is ~21 KB of JSON a chunk, and
-  the text and vector go up again (~11 KB).
+- Done (2026-10-29): embedding on the gateway's side of a document PUT.
+  `llmgw-kb` sends text only to a gateway that says `server_embeds`; the
+  gateway embeds through its own `/v1/embeddings` as the caller, 32 a batch,
+  four batches at once, waiting out a busy engine. On 40 DTIC files the
+  upload went from 25.7 to 3.5 MiB, and the ~21 KB a chunk that came down
+  is gone; on 6 files both ways gave the same 32 chunks and the same search
+  answer. Speed is the engine's either way: the PoC's bge-m3 on node02's two
+  CPU cores embeds ~0.2 chunks/s, so a real import still wants it on a GPU.
 - The store next: HNSW inserts took half of one of pg01's two cores; at a
   million chunks the vectors alone outgrow its 4 GB. halfvec (pgvector ≥
   0.7) and a larger store machine.
@@ -1426,13 +1493,660 @@ console show and add them.
   `on_demand.evict_after_idle` / `min_lifetime` as live settings; an
   upgrade that leaves nodes behind says which; `ctl model serve` serves an
   embedding model for embedding; refreshes wait while a form is filled in.
-- Open: llmgw's `/mcp` with picture, 3D, VLM and workflow tools, so a
-  client needs no server of its own (asked by the game team); the shared
-  ComfyUI replica (H3's placement side).
+- Done 2026-10-15: llmgw's `/mcp` with picture, 3D, VLM and workflow tools,
+  so a client needs no server of its own (asked by the game team); the
+  shared ComfyUI replica (H3's placement side).
+- Done 2026-10-16 (v0.14.6, v0.14.7): setup lists every part with what it
+  is and its settings; three machines each run the control plane, with a
+  floating address (VRRP id from the address); a lost store primary costs no
+  request (3 s connect per member, keepalive and TCP_USER_TIMEOUT, Patroni
+  ttl 20, a standby leaves the lease alone a round after the store returns);
+  a new leader waits for its judge within the budget; a judge is asked again
+  on another machine at half its budget; init sizes the judge (2 GB, no
+  prompt cache, two cores, one judge on up to four cores on one machine).
+
+## I. What other gateways do that llmgw does not (written 2026-10-16)
+
+From a comparison with Bifrost (maximhq/bifrost, Apache 2.0, Go), with
+LiteLLM and Portkey in mind. Bifrost routes to 23+ cloud providers and runs
+no model; llmgw runs the models, so most of Bifrost is either in llmgw
+already (keys, budgets, fallbacks, guardrails, tracing, SSO, HA — the last
+three paid in Bifrost) or does not apply. These four are worth taking for
+a site that runs its own models. Left until a customer asks: native
+Azure/Anthropic/Bedrock/Vertex adapters, request hooks a customer plugs in,
+routing by what a request holds. Not taken: a realtime (WebSocket audio)
+API, prompt management, an `npx` install.
+
+### I1. Answers to similar questions, not only the same one (M) — done 2026-10-25
+`keys cache --similar`, `router.cache.embedding_model`, `internal/router/similar.go`. Changed
+from the plan: the vectors are in the gateway's memory beside the exact answers, not a pgvector
+table. The exact cache keeps answer text out of the store on purpose (data.md), and a table of
+answers would undo that; it also works on SQLite. The usage row marks a similar hit as a hit
+(no column of its own). Open: the 40-rewordings and eval-suite runs on the PoC.
+The original item follows.
+The answer cache (`internal/router/cache.go`, C17) answers a request whose
+body is the same byte for byte. A help desk asked "how do I reset my
+password" in forty wordings asks the model forty times.
+- Per project, off by default, beside the exact cache: `llmgw keys cache
+  --project P --similar 0.95` (the cosine similarity a question must reach;
+  higher is stricter). The exact cache is asked first.
+- The question embedded by a model the operator names
+  (`router.cache.embedding_model`, an embedding model in the catalog), into a
+  pgvector table per project with the answer, the model and the TTL; the
+  store's halfvec and text index work from F applies. SQLite stores keep the
+  exact cache only.
+- What counts as the question: the last user turn, with the system prompt
+  and the earlier turns hashed and required to match exactly — a similar
+  question in another conversation is not the same question. Tools,
+  `response_format` and the sampling settings also exact.
+- Said: `x-llmgw-cache: similar` with the similarity, and the usage row
+  says so; guardrails judge the request before a hit, as for the exact
+  cache. A project with `no_store` gets neither.
+- Done when: forty rewordings of one question on the PoC cost one model call
+  and 39 lookups, and an eval suite run with the cache on scores within a
+  point of the run with it off.
+
+### I2. MCP gateway: the customer's tool servers behind llmgw (L)
+P1 done 2026-10-20: HTTP servers registered (`ctl mcp`, `/api/v1/mcp/servers`,
+store `mcp_servers`), tools as `server.tool` in `/mcp`, `allowed_tools`
+grants in the tree (deny by default), credentials in the store or by
+reference, the caller named in `X-Llmgw-*` headers, time and size bounds,
+progress passed on, audit (Source `mcp`) and metrics. Left: P2, the
+guardrails both ways, a per-tool rate limit, a breaker, and a portal page;
+P3, stdio servers and per-person OAuth. The "done when" with Claude Code
+against a real server is still to try.
+`/mcp` serves llmgw's own tools. An agent inside a company also wants Jira,
+the file server, a database — each an MCP server, each today configured
+in every client with its own credentials and nobody seeing what was called.
+- Servers registered once, by an administrator: `llmgw ctl mcp add jira
+  --url https://… --header-file /etc/llmgw/secrets/jira` (HTTP transport;
+  stdio servers run on the control plane under their own user, like an
+  engine). Their credentials stay on the control plane, never in a client.
+- `/mcp` lists every server's tools, prefixed by its name (`jira.search`),
+  filtered by what the key may use: an allowlist per org, team, project or
+  key, by server or by tool, in the tenancy tree as model allowlists are.
+- Each call goes to the audit log — who, which tool, the arguments' size,
+  the outcome and how long — and through the guardrails both ways (a tool's
+  answer is prompt input). A per-tool rate limit.
+- Later, if it earns it: Bifrost's "code mode" (the tools as one typed API
+  the model writes code against), which cuts the tokens of a long tool list.
+- Done when: Claude Code given only llmgw's `/mcp` reaches a test MCP server
+  through it, a key without the grant does not see the server's tools, and
+  every call is in `ctl audit`.
+
+### I3. What llmgw adds to a request, measured (S) — done 2026-10-16
+`llmgw ctl bench overhead` (`internal/cli/bench_overhead.go`): an instant
+engine and the gateway in one process, four modes (direct, bare, auth, auth
+and guard patterns) at rising rates. On an Apple M5 the gateway added
+0.05–0.42 ms at p50 and at most 0.49 ms at p99, in the README.
+`TestTheGatewaysOverheadIsMeasured` runs a second of it in CI with no bound
+asserted. Left: the CPU figure is rough with everything in one process; a
+gateway in a process of its own would measure it properly. The original
+item follows.
+
+Bifrost's headline is "11 µs at 5k RPS" (its own figure, and its own pages
+disagree with it). llmgw has none, so any question about its overhead is
+answered with a guess.
+- `llmgw ctl bench --overhead`: the same load through the gateway and straight
+  at a `sim-engine` (`internal/cli/simengine.go`) that answers at once, at
+  rising request rates, with auth, accounting and the guardrails' pattern
+  checks on and off. Reports the added latency at p50/p99 and the gateway's
+  CPU per thousand requests.
+- On the PoC the CPU is oversubscribed (pve1, 28 threads for ~125 vCPUs):
+  report it as a ratio and record the host's load beside it, and repeat on
+  an idle machine before a figure goes into the README.
+- Done when: the README states the overhead, with how it was measured, and
+  CI runs the short version so a change that doubles it is seen.
+
+### I4. Several keys for one outside provider (S) — done 2026-10-17
+`external.api_key_files` (`catalog.ExternalKey`: file, weight), one endpoint
+each in `control.externalEndpoints`, 401 retried on another key and
+`external_key_refused` notified (`router.keyRefused`). With one key a 401 is
+still the client's. `TestAKeyAtItsLimitOrRefusedIsPassedOverForAnother`. The
+original item follows.
+
+An external model has one `api_key_file` (`internal/control/control.go`,
+where external endpoints are made). A provider's limit is per key; one key
+answering 429 makes the whole fallback answer 429.
+- `api_key_files: [a, b, c]`, each with an optional `weight`: one endpoint
+  per key (`external:<model>#<n>`), so the router's scoring, retries and
+  breakers work per key as they already do per replica. A key answering 429
+  or 401 is passed over while its breaker is open; a 401 also notifies
+  (`external_key_refused`).
+- The usage row says which key answered (by its index, never the key).
+- Done when: with two keys and one returning 429, every request is answered
+  by the other, and the console shows the one breaker open.
+
+## J. Found testing HA and single-machine installs (written 2026-10-16)
+
+Three VMs installed as HA from the setup screens, a fourth as a judge node,
+a fifth as a single 4-core, 8 GB machine; machines powered off under load.
+
+### J1. A restart refused for room another restart was giving back (S) — done 2026-10-16
+Not a wait for ever, as first written: on the 4-core VM the judge's restart
+was *refused* ("fits on no node even with this replica stopped") because the
+starter, restarted at the same time to reserve fewer cores, still held its
+old ones. Once refused, nothing tried again, so the judges kept their old
+figures. Two fixes in `internal/control/reconciler`:
+- a restart whose replacement fits once other restarts under way have gone
+  (queued, retiring or draining) waits for them, up to DrainGrace, instead
+  of being refused (`fitsAfterRestarts`, restart.go);
+- a drained replica counts a retiring one that still answers as serving, so
+  one restart after another no longer leaves the first's drained replica,
+  and the room it holds, waiting out its grace (reconciler.go).
+`TestARestartWaitsForRoomAnotherRestartWillGiveBack` is the 4-core case.
+
+### J2. A judge the leader adds is sized as init's is (S)
+When a policy names a judge the catalog lacks, the leader serves it itself
+(`internal/cli/judgekeeper.go` via `knownJudge.serveRequest` in
+`judges_known.go`, through modelserve). That judge reserves 3 GB (the
+weights, an 8192-token KV cache and a fixed gigabyte), no cores, and runs
+with llama.cpp's prompt cache, which grew one to 3.2 GB. init's judge
+(`guardModel` in `internal/cli/serverinit_cluster.go`) reserves 2 GB and
+its cores and runs with `--cache-ram 0`.
+- `modelserve` gives a guard model `--cache-ram 0`, and the keeper's
+  request the same memory and cores rule as `cpuPlan`.
+- Done when: a judge removed from the catalog comes back with the same
+  figures init would have given it.
+
+### J3. HA setup puts the judges on nodes of their own (S)
+Under load on a 4-core control plane that was also a node, 38 of 60
+requests were refused for want of a judgment. With the judges on a node of
+their own (4 cores, 32 GB, two cores each) the same test refused 2.
+- The HA setup offers a judge node or two: machines given like the store's,
+  added as nodes with a label (`llmgw.role: judge`), and the judge's
+  profile selects that label.
+- Done when: the setup installs the judges on the machines given for them,
+  and the load test above refuses no more than it did by hand.
+
+### J4. The model mirror on more than one machine (M) — done 2026-10-29, tried on the PoC
+In an HA install the mirror stays on the first machine. While it is down,
+models the nodes already have keep serving and new ones wait for it.
+- A second mirror kept in step (each machine of the store, or one more), the
+  control planes and nodes given both, the first that answers used.
+- Done when: with the first machine powered off, a model added then is
+  fetched and served.
+
+### J5. An upgrade corrects an earlier install's judge (S) — done 2026-10-16
+Built in the judge keeper (`internal/cli/judgekeeper.go`, `refit`), on the
+leader: a judge whose entry is exactly what init before v0.14.7 wrote is
+given 2 GB, `--cache-ram 0` and cores by init's rule (`refitCores`: up to
+four for one judge, two each for more), through `modelsvc.SetProfiles`,
+and its ready replicas are restarted (after J1). Its replica count is left
+as it is. One that differs is left alone and said once in the log. The
+original item follows.
+
+v0.14.6 and earlier installed two judges reserving no cores or one, with
+2 or 4 GB and the prompt cache on. On fewer than eight cores every judgment
+missed its 5 s budget and every request was refused. v0.14.7's notes say
+how to fix it by hand (`ctl model resources`, then `ctl restart`).
+- On upgrade, or on `server init` run again, a judge whose profile is still
+  exactly what an earlier init wrote is moved to today's figures and
+  restarted (after J1). One changed by hand is left alone and named.
+- Done when: a v0.14.6 single-machine install upgraded to it judges within
+  budget without anything typed.
+
+### J6. Rate limits per process (note)
+Rate limits are counted in each control plane's memory. A failover starts
+every key's minute afresh. Documented in install.md; worth a shared count in
+the store only if a customer's limits must hold across a failover.
+With `router.active_active` (O2, 2026-11-01) the counts are shared every 2 s,
+so a standby taking over keeps the minute; without it, as above.
+
+### J7. nginx in front of several control planes (S)
+`deploy/proxy/Caddyfile` sends to whichever control plane answers `/leader`.
+nginx's open-source build has no active health checks, so its `upstream`
+block would send to standbys and to a lost leader alike.
+- nginx.conf: the floating address as its one upstream, and why; or the
+  `max_fails` passive approach with what it costs.
+
+### J8. A judge budget that fits the machine (S)
+A judgment takes 2.3 s on four cores and 4.2 on two, against the 5 s
+default: two cores are close to it, and a busy machine passes it.
+- At install, after the judge is up, time three judgments. If the slowest is
+  more than half the budget, say so in the plan's closing text with the
+  choices (more cores, a GPU judge, a longer `latency_budget`, `fail:
+  open`). The budget itself is not changed without being asked.
+- Done when: an install on a machine too slow for its judge says so before
+  its first refused request.
+
+## L. Taken from Kong's AI Gateway (written 2026-10-25)
+
+From a look at Kong AI Gateway 3.9–3.16 and 2.x. Kong's open source stops at
+3.9, and only six of its AI plugins are open (proxy, prompt template,
+decorator, regex prompt guard, request and response transformer); the rest
+is Enterprise or Konnect. So nothing of Kong is embedded: the two meet over
+plain OpenAI/Anthropic HTTP, and these are the ideas worth taking. Not taken:
+prompt compression (Kong ships it as a private LLMLingua sidecar, its benefit
+unshown), prompt templates and decorators (no one has asked), Konnect's
+entity model. MCP items are llmgw-80's (I2); they are listed for it.
+
+### L1. Traces in the OpenTelemetry GenAI conventions (S) — done 2026-10-25
+`gen_ai.*` on the `inference` and `upstream` spans (`internal/router/trace.go`):
+operation, request and response model, max_tokens, temperature, top_p, the
+session as conversation id, input/output/cached tokens, the engine as
+provider. Never prompt or answer text (`TestTracesSpeakTheGenAIConventionsAndHoldNoText`).
+Prometheus already counted tokens by direction, cached apart. Left: the MCP
+calls' spans as `gen_ai.tool.*`, once /mcp is traced at all.
+
+### L2. Redacted personal data put back in the answer (S) — done 2026-10-25
+`restore: true` (`guard.Rules.Restore`), `Verdict.Restore`, `internal/router/restore.go`
+ahead of the output filter. The original item follows.
+The guardrails rewrite personal data as `[KIND_n]` (`internal/guard/gate.go`,
+`placeholders`), the same value always the same placeholder. A model that
+answers "Dear [PERSON_1]" leaves it so. Kong's sanitizer restores.
+- Per policy, `restore: true`: the placeholders of one request kept in memory
+  for its life, and swapped back in the answer, streamed too (a placeholder
+  split across two chunks held until it is whole).
+- Only into the answer to the request that held them; never stored, never in
+  usage rows or traces.
+- Done when: "write to Ayşe Yılmaz at ayse@example.com" through an enforce
+  policy reaches the model as placeholders and the client reads the real name
+  and address back.
+
+### L3. A guardrail that is someone else's service (S) — done 2026-10-25
+`guard.Service` (`internal/guard/service.go`), policy `services`, asked
+before the judge on input and once on the whole answer. The original item follows.
+Customers may be required to use one: Azure Content Safety, Lakera, NeMo
+Guardrails, an in-house classifier. Kong's AI Custom Guardrail is a
+templated HTTP call with an expression for "block".
+- A detector kind `http`: URL, headers from secret files, a body template
+  with the text, a JSON path that says blocked and one that says why;
+  timeout and `fail` as the judge's. Checked on input, output or both.
+- Done when: a policy with an `http` detector against a test service blocks
+  what it says to, and a service that does not answer is treated as `fail`
+  says.
+
+### L4. Routing by latency and by cost, and failover by priority (M) — latency done 2026-10-25
+`endpoint.Registry.ObserveTTFT`, `scheduler.Latency()` in `chat`. Failover in tiers is
+what `fallbacks` already does (local, then an outside model when no room). Left:
+routing by cost between several outside providers. The original item follows.
+The scorers (`internal/router/scheduler/scorers.go`) know sessions, queues
+and KV cache. Kong also balances by measured latency, by tokens or cost
+used, and in priority tiers.
+- A `latency` scorer: a moving average of time to first token per endpoint.
+- Fallbacks in tiers: local replicas first, then an outside provider, with
+  the tier said in the usage row.
+- Done when: with two replicas of one model, one on a slower card, `latency`
+  sends most of the traffic to the faster and the slower still gets some.
+
+### L5. Prices by kind and by cache (S–M) — done 2026-10-25
+`cached_per_1k`, charged on the answer's `cached_tokens`. A budget in money was already
+there (`keys limits --budget`). Not done: cache-write and audio/video token prices —
+no engine llmgw runs reports them, and no provider it reaches is billed by them yet.
+The original item follows.
+`catalog` prices are per million input/output tokens and per picture/mesh.
+Kong prices cached reads and cache writes apart, and audio, image and video
+tokens apart, and limits by cost.
+- Price fields for cached input and by modality; the usage row's cost from
+  them; a budget limit in money as well as tokens (`keys limits --cost`).
+- Done when: a request with 90% cached prompt costs what the cached rate says.
+
+### L6. REST APIs as MCP tools (M) — for llmgw-80, beside I2
+Kong turns OpenAPI operations into MCP tools. A company with an internal
+REST API and no MCP server could offer it to agents through `/mcp`:
+`llmgw ctl mcp add hr --openapi https://hr.internal/openapi.json`, each
+operation a tool, under the same grants, audit, guardrails and rate as I2.
+
+### L7. MCP sign-in by the MCP spec (M) — for llmgw-80, part of I2 P3
+Kong's MCP OAuth2: llmgw's `/mcp` as an OAuth 2.1 resource server (protected
+resource metadata, tokens checked by JWKS against the company's identity
+provider), and a token exchanged for a short-lived one per upstream server
+rather than one long-lived secret.
+
+### L8. A2A (S, when asked)
+Agent-to-agent traffic through the gateway: agent cards' URLs rewritten to
+llmgw's, calls audited and traced. The standard is young; wait for a
+customer.
+
+### L9. llmgw behind Kong, written down (S) — done 2026-10-25
+docs/kong.md, tried with kong:3.9: plain route and ai-proxy, keys passed on;
+Kong 3.9 counts no completion tokens of a stream. The original item follows.
+A customer that already runs Kong puts llmgw behind it: `ai-proxy` with
+`provider: openai` and `upstream_url` at llmgw (3.9, open source), or a plain
+route. A page in docs/ with the configuration, tried against Kong 3.9 in a
+container, and what Kong's token accounting sees.
+
+## M. Models from Hugging Face we can trust (written 2026-10-25)
+
+The user asked how we know a model fetched from Hugging Face is safe. Two
+questions: whether its files can run code on our machines, which we can
+close, and whether the model itself was trained to misbehave, which no scan
+proves — only who published it, an eval before it serves, and the
+guardrails narrow that. What existed: `--trust-remote-code` refused, no
+repository Python fetched, ComfyUI nodes pinned to a commit, engines
+hash-pinned and run as `llmgw-engine`, mirror versions pinned to a commit
+and never overwritten. What did not: everything below.
+
+### M1. Weights in a pickle format refused (S) — done 2026-10-25
+`.bin`, `.pt`, `.pth`, `.ckpt`, `.pkl` run code as they are loaded, which is
+how the malicious models found on Hugging Face work. The node refuses such a
+source, or a mirror set holding one, before fetching anything; a model being
+added or given such a source is refused; vLLM and SGLang downloading a model
+by its id are given `--load-format safetensors`. The explorer hides a repository with neither GGUF nor safetensors
+(under "runs here", ticked by default) and does not offer a repository's
+pickle files. `allow_pickle: true` on a model lets it through, on purpose. On the
+PoC, `triposr` (`hf:stabilityai/TripoSR/model.ckpt`) is the one model that
+needs it.
+
+### M2. Every download pinned and checked against Hugging Face (M) — done 2026-10-30
+Done: what is added from Hugging Face is pinned (the file's commit URL, or
+`--revision` for vLLM/SGLang) with each file's sha256 in the origin; the node
+refuses and deletes a copy that is not it, naming both, and keeps the
+checksum beside a cached file; the mirror checks what `ctl model serve`
+fetches, `mirror add` a Hugging Face URL. Left: models declared in the
+settings are not pinned, the admin page's and the console's mirror fetches
+send no sha256, and adapters by hub id are not pinned.
+- `hf:org/repo/file` follows `main` and is checked by nothing; a bare
+  `org/repo` given to vLLM is downloaded by vLLM, of whatever `main` is.
+  Both: resolve to the commit when the model is added, keep it in the
+  catalog's origin, and fetch that commit only (`--revision` for vLLM).
+- The sha256 Hugging Face publishes for each LFS file compared on every
+  path, and a file that differs removed: `mirror add`, the manager, the
+  node's `hf:` and http sources, mirror set files with no sha recorded.
+- Done when: a file changed on Hugging Face after a model was added is
+  refused by the node with both checksums named.
+
+### M3. Hugging Face's own scan (S)
+Hugging Face scans every file (pickle imports, ClamAV, partners' scanners)
+and marks a file unsafe. Ask for it with the repository listing; refuse an
+unsafe file, show the result on the mirror page and in the catalog's origin.
+The API field has to be confirmed first.
+
+### M4. Publishers we trust, and the rest asked about (S)
+An allow-list of organisations in the cluster's settings (Qwen, meta-llama,
+mistralai, google, microsoft, …). A model from anyone else is added only by
+an admin who says so, and who said it is kept in the origin. The PoC already
+serves four community models whose training nobody here checked.
+
+### M5. Engines without the internet (M)
+The firewall is inbound only, so an engine that was made to run code can
+send what it reads anywhere. With weights from the mirror the engines need
+no egress: an nft output rule for the engine user (allow the mirror, the
+cluster, loopback), systemd hardening of the agent's engines
+(`NoNewPrivileges`, `ProtectSystem`, private /tmp). Agent itself still root.
+
+### M6. An eval before a new model serves (S)
+`eval` and `canary` exist; make a new model from an untrusted publisher
+(M4) serve only after its eval suite passes, so a model trained to answer
+wrongly on a trigger has at least been asked the suite.
+
+## N. Strata, all the way up (written 2026-10-30)
+
+Strata (opt-in engine since 2026-10-29) is wired at the agent: adapter, `llmgw strata-serve`, install pinned by
+digest, metrics in vLLM's names, GPU memory of child processes. Above the agent it is invisible: a Strata model is
+served only from a definition written by hand. In order:
+
+**A. What is broken.**
+- N1. A Strata named by hand (`engines.strata.dir`) was never reported: no Probe. Done 2026-10-30.
+- N2. Its engine runs in a session of its own and survived a stop, holding the card and the RAM: the supervisor
+  now finds what left the group before signalling it. Done 2026-10-30.
+- N3. Host memory: the experts live in RAM (40 GB for Qwen3.8-Flash-Next IQ2_XS, 72–136 GB for Unsloth's) and
+  nothing reserved it. A Strata profile added now reserves what its size takes, from the table (N5), unless it says
+  `min_memory_gb`. Done 2026-10-30.
+- N4. Pascal and Volta: installed (CUDA 12.9 build) and then refused (7.0 floor). Floor now 6.0. Done 2026-10-30.
+
+**B. Offered where models are chosen.**
+- N5. A table of Strata's families and sizes (qwen, coder, swift, unsloth, gyro, nvfp4, qwen36, deepseek41) with
+  download and RAM, from Strata 0.2.7's setup (internal/strata/models.go); the catalog refuses an unknown family or
+  size, and requires `--size`. Done 2026-10-30. Speeds measured: Qwen3.8-Flash-Next IQ2_XS 70–87 tok/s on an RTX 6000.
+- N6. `modelplan`, `ctl model plan` and `ctl model serve` offer "Strata on one card + N GB RAM" for a model it runs;
+  a large MoE that fits no card is not "fits nowhere". Done 2026-10-30, without a speed (one measurement so far).
+- N7. The web serve page, the explorer's fit, the console's serve form: the same choice. Done 2026-10-30.
+- N8. `--context`/`--parallel` tied to `context_length`/`max_output`; `ctl model context` takes Strata.
+  `ctl model context` takes Strata (setup's `--context`), and `--max-output` changes the longest answer alone,
+  in place, with no restart: done 2026-11-01 (31f440a). Open: a Strata model served without `--context` still runs
+  at what Strata's setup picks (4096 on cuda01) while the catalog says the model's trained context (32768), so
+  serve should write `--context` from `context_length`; and `--parallel`.
+
+- N7b. The mirror page's Strata tab: every family and size, fetch (GGUF) or serve (packs Strata fetches itself:
+  NVFP4, DeepSeek V4.1). Done 2026-10-30. Community `strata-ple` packs Strata 0.2.7 does not list stay unservable;
+  running them would mean starting the engine past Strata's setup (`--pack`, `--native`, `--ple-gguf`).
+
+**C. Operating it.**
+- N9. `ctl engine-check strata`: the flags of setup.py and server.py, which drift between versions.
+- N10. `ctl engine-build NODE strata cuda12|cuda13`, as vLLM has.
+- N11a. Strata's per-replica state directories (engine-home/strata/<replica>) are left behind when a replica stops;
+  remove them with it.
+- N11. Disk for the prepared pack and draft layer (and DeepSeek V4.1's 240–280 GB), cleaned with the weights; a
+  first start's preparation within the startup timeout.
+- N12. docs/install.md: Strata in the engines table, a section with a profile, its memory and first-run time;
+  `doctor` and `server init` checking or offering it.
+
+**D. What Strata can do that llmgw does not ask of it yet.**
+- N13. Several cards of one node (`--gpus`, `--gpu-roles`): larger packs on a customer's multi-GPU machines.
+- N14. `--calibrate` once per node and model, kept.
+- N15. Vision, KV types, experts from SSD when RAM is short (`--low-ram`) in the plan and forms.
+- N16. Versions apart from llmgw's releases (Strata's signed update feed, `--rollback-engine`); offline bundles.
+- N17. `cache_salt` for KB isolation if Strata caches prefixes; `ctl bench` with its metrics path.
+
+## O. More than one place answering clients (written 2026-10-30)
+
+Every client request goes through the leader's router today: the floating address is on the leader, and a standby
+passes `/v1` on to it (a standby's router sees an empty cluster). That is failover, not scale: the cluster serves what
+one control plane's router can. In order:
+
+- O1. Measured 2026-10-30/31 on the PoC: the leader (2 vCPU, 2 GB, pve1 at load 6–10 of 28 threads) in front of
+  a fake engine on node06 (`sim-engine`, 500 tok/s a stream, no batch slowdown, as an external model), 64-token
+  streams from mirror 146, the default `observe` guardrails and usage written to PostgreSQL:
+
+  | at once | req/s | tok/s | first token p50/p95 | leader llmgw CPU |
+  |---|---|---|---|---|
+  | 64 | 465 | 29,752 | 9 / 34 ms | 159% |
+  | 128 | 572 | 36,607 | 91 / 167 ms | 181% |
+  | 256 | 577 | 36,900 | 310 / 392 ms | 182% |
+  | 512 | 551 | 35,240 | 793 / 883 ms | 181% |
+
+  The ceiling is the router's CPU: ~575 requests/s, ~37k streamed tokens/s on two cores, no errors; past it the
+  first token waits. Found on the way and fixed: an external model held to 3–4 at once by admission; a valid key
+  throttled as "unrecognised" when its cache entry expired under load.
+
+  Profiled under 128 streams (2026-10-31): a third of the CPU was one write a token to the client; joining the
+  events already in hand into one write took the ceiling to ~700 req/s and ~45k tok/s, first token p50 45 ms at 128.
+  Now the largest part is the output guardrail of the default `observe` policy, ~30%: decoding each event (~20%;
+  a token-level reader with encoding/json/jsontext was measured slower than json.Unmarshal, 1.66 vs 1.27 µs) and
+  the PII patterns (~12%). Ways on, a guardrail decision: judge shadow output in larger windows, or a sample of it.
+  Next: the same run on a machine the size of a customer's control plane, then O2.
+- O2. Every control plane answers `/v1` itself (active-active):
+  - [x] the leader's replica table followed live by the standbys, so their routers route (2026-11-01, behind
+    `router.active_active`, off by default). Every 2 s a standby posts what its router used to the leader's
+    `POST /api/v1/routing/activity` over the agent port with its control-plane certificate, and is answered with
+    the Ready replicas: no keys travel, each control plane derives a replica's key from the cluster secret. The
+    leader takes the last uses in (else a replica busy behind a standby is evicted as idle) and the on-demand models
+    asked for; a catalog changed on the leader is read again from the store. Unheard from the leader for a minute,
+    a standby routes nothing and passes `/v1` on as before. The portal, `/admin` and `/mcp` still go to the leader.
+  - [x] budgets: each router raises its counters to the store's every 30 s, so another's spend is seen within that.
+  - [x] rate limits cluster-wide (2026-11-01): each router's per-counter request and token counts, this minute's
+    and the last as absolute figures, ride the same 2 s exchange; the leader answers each standby with everyone
+    else's sum, its own included, and puts the standbys' sum in force on its own router. Each limiter adds the
+    others' counts to its sliding window. Overshoot is bounded by what N routers allow in 2 s; a standby quiet for
+    10 s stops counting. (J6's single-process case was never the problem.)
+  - [ ] a session's requests to one router and one replica: the session ring already picks the same replica on
+    every router for the same replica set; what is left is the same router, which is the balancer's (source hash).
+  - [ ] in front: IPVS on the keepalived the setup already installs, DNS round-robin, or the customer's balancer
+    (J7, L9). Until then the standbys' own addresses are what spread the load.
+  - [x] tried on the PoC (2026-11-01, v0.16.1-14-g8b7eefe, setting on since): 140 leads, 150 follows. A request
+    through 150 is answered and logged by 150; a model added on the leader is routable on 150 within a tick; a
+    CPU replica used only through 150 shows idle 6 s on the leader. Load from 146, 64-token streams, a sim engine on
+    node06 at 5,000 tok/s a stream (2,300 req/s on its own; at 500 tok/s it was the ceiling, ~490 req/s, routers or
+    not), pve1 at load 14 of 28 threads:
+
+    | streams | leader alone | split over 140 + 150 | first token p50, alone / split |
+    |---|---|---|---|
+    | 128 | 971 req/s (140 at 176%) | 832 + 692 = 1,524 req/s (178% / 166%) | 115 / 59–75 ms |
+    | 256 | 990 req/s (177%) | 816 + 677 = 1,493 req/s (177% / 164%) | 243 / 135–166 ms |
+
+    1.5× with two routers on an oversubscribed host, no errors. Not 2×: both routers' VMs and the load generator
+    share pve1's cores.
+  - Known and accepted for now, per router: the precise KV index is empty on a standby (agents stream to the leader
+    only), a revoked key works on a standby until `auth.cache_ttl`, canary promotion sees the leader's share only,
+    the answer cache and prefix tree are split, an MCP caller's rate is per router.
+- O3. Only if O2 is not enough: stateless gateway machines (`llmgw server --role gateway`) in front of the leader's
+  control plane, as many as the traffic needs.
+
+## P. Training beyond one card (written 2026-11-11) — decision open
+
+F4 trains a LoRA on one card of a node the caller names (`ctl train start --node`), one job per node, no resume.
+Larger work (full fine-tuning, a LoRA on a 70B, several GPUs or machines) needs more. Two ways:
+
+**A. In llmgw.** Add, in order: a job that takes N whole GPUs on one node (torchrun, FSDP/DeepSpeed in the agent's
+venv); placement by the control plane instead of `--node`, freeing GPUs from replicas by the scaling schedule or by
+draining; checkpoints to shared storage and resume when an agent or node goes; datasets uploaded or read from
+storage, not only KB pairs; then gang placement of N nodes with a rendezvous on the first, the NCCL environment from
+the fabric discovery (IB/NVLink, `llmgw doctor`); a console screen with logs and loss. Size: L for one node, XL for
+several. For: one product, one install, one set of users and grants (KB rules carry over), GPUs shared with
+inference without a second scheduler. Against: it is Slurm's gang scheduling, requeue and accounting written again;
+multi-node NCCL cannot be tried on the PoC (one GPU node, no IB); every customer's training code would have to fit
+our job format.
+
+**B. Slurm beside llmgw.** The TUI installs Slurm on the nodes the customer picks (munge, slurmctld on the control
+planes, slurmd, users from the same LDAP/AD), with pyxis/enroot for containers. A node is handed between the two:
+llmgw drains its replicas and `scontrol update state=resume`, and back. llmgw's part is what comes out: a checkpoint
+or adapter from a Slurm job published to the mirror (`ctl train import`), served through canary and the eval. Size:
+M. For: what research and university customers already run (`sbatch`, `srun torchrun`), any training code, mature
+multi-node and requeue; we write only the install and the handover. Against: a second system with its own failures,
+upgrades and users to keep in step; GPUs split by hand or by the handover, not by demand; access rules on training
+data are Slurm's and the file system's, not the KB's.
+
+Both need shared storage (NFS, BeeGFS, Lustre) the customer provides; neither changes F4.
+
+Recommendation: keep F4 in llmgw for small single-card adapters tied to KB grants; take B for anything larger;
+build only A's first step (N GPUs on one node, placed by the control plane, with resume) if customers ask for
+mid-size jobs without wanting Slurm.
+
+**B taken (2026-11-11).** Design: the cluster's settings say it (`slurm: enabled, cluster_name, controller, nodes,
+lent, partition`); the leader tells each named node's agent (root) what to be, the agents install and run it.
+- The controller is an llmgw node (its agent runs slurmctld); `nodes` run slurmd configless
+  (`--conf-server`), so slurm.conf, gres.conf and cgroup.conf live only on the controller.
+- The munge key is derived from the cluster secret (HKDF), so nothing new is stored or copied by hand.
+- Each node reports `slurmd -C`; the leader writes slurm.conf from those lines, its mgmt address and its GPUs
+  (`Gres=gpu:<type>:N`, gres.conf with the device files), and sends it to the controller again when one changes.
+- A node in `nodes` belongs to llmgw until it is in `lent`: drained in Slurm (reason "llmgw: serving models").
+  Lent: llmgw drains it as `ctl drain` does (kept in force each tick, as facility holds are, so a new leader
+  keeps it) and, once no replica is left, its agent resumes it in Slurm. Taken out of `lent`: the agent drains
+  it in Slurm, reports the jobs left, and when none are the leader makes it ready for llmgw again.
+- Phases: P1 the `slurm` package (conf, key, parsers) and settings; P2 the agent side (proto, install, munge,
+  slurmd/slurmctld, status); P3 the leader side (conf from reports, lend/reclaim); P4 `ctl slurm`
+  (status, lend, reclaim) and a TUI screen; P5 `ctl train import` of a Slurm job's output into the mirror.
+  Later: a backup slurmctld on a second node, accounting (slurmdbd; it takes MySQL/MariaDB only, not the Patroni PG), users from LDAP/AD.
+- Tried on the PoC 2026-11-11 (dev build v0.16.3-17-g849760c-slurm on 140, 150, node05, node06; Ubuntu
+  24.04, Slurm 23.11.4 from the archive): `ctl slurm setup --controller llmgw-node06 --nodes llmgw-node05`
+  installed munge, slurmctld and slurmd, node05 entered drained with its slurmd -C hardware (8 CPUs,
+  31132 MiB after the 3%). `lend`: its 2 replicas moved off, then IDLE in Slurm; srun and sbatch ran there.
+  `reclaim` with a 90 s job running: MIXED+DRAIN, llmgw kept off, ready again within a minute of the job
+  ending. Found and fixed on the way: a client command needs its files even for --version (version now
+  from the daemon); the packaged unit is Type=notify (slurmd needs --systemd); slurmd's own "slurmstepd
+  infinity" was counted as a job; without accounting storage 23.11's main scheduler marks batch jobs
+  InvalidAccount and backfill starts them, one a pass (bf_interval=10: 6-10 s each). So many small jobs
+  want slurmdbd. — Wrong cause, found 2026-11-11: a job asking no memory got the whole node's (ReqTRES
+  mem=all), so a node ran one job at a time and each next one waited for backfill's pass. Fixed: the
+  partition's DefMemPerCPU is the leanest node's share; slurmdbd is not needed for this. InvalidAccount is real too:
+  23.11 dropped accounting_storage/none, and without slurmdbd the main scheduler passes every job by;
+  backfill starts it within bf_interval (10 s). priority/basic did not change that (fair share needs
+  slurmdbd anyway). Starting on submission needs slurmdbd, on MySQL/MariaDB.
+- slurmdbd, done 2026-11-12 (748f777): the controller's agent installs MariaDB (bound to 127.0.0.1; a
+  MariaDB that was there keeps its settings), makes `slurm_acct_db` and its user (password from the munge
+  key, through a 0600 file, never argv), runs slurmdbd (6819, in the firewall for the cluster) and adds
+  the cluster, before slurmctld. slurm.conf: AccountingStorageType=slurmdbd, jobacct_gather/cgroup,
+  AccountingStorageTRES=gres/gpu, priority/multifactor. Tried on the PoC: five jobs started on
+  submission, `sacct -a` has them with their TRES. Not tried: a fresh `server init --slurm` (needs VMs);
+  a Red Hat controller.
+- Fresh install tried 2026-11-12 on VMs (3 control planes with Patroni, 2 CPU nodes, `server init
+  --slurm`): Slurm, slurmdbd and MariaDB came up with the install; two-node srun/sbatch on the lent nodes,
+  `sacct` has them; leader failover kept serving. Open from it: `--slurm` needs the first machine to be a
+  node (the controller is its node), so a site wanting control planes without nodes chooses the
+  controller later with `ctl slurm setup`; batch output lands on the job's first node until there is
+  shared storage; on shared CPUs the judge refuses the first request after a placement change (fail
+  closed, ~10 s), and its advice repeats `--cores 4` to a judge that has 4. Tried again on the idle
+  VMs: one judge restarted in place, then both at once, requests every second: 48 of 48 served; the
+  first after a cold start took 7.6–8.9 s end to end against ~3 s, within the 5 s judge budget. So the
+  refusal was the install's load on the CPUs; if it is seen again, warm a new judge with one judgment
+  before it takes requests.
+- Shared files and users, built 2026-11-12: `slurm.shared` {path, nfs | server} — the site's export
+  mounted (a systemd mount unit, hard, nconnect=4), an llmgw node exporting the path itself
+  (/etc/exports.d, to the Slurm machines' addresses, root_squash; 2049 in the firewall), or the site's
+  own mount only looked for; `slurm.users: llmgw` — every active account a Linux user everywhere
+  (`slurm_users`: login from the e-mail, UID from 200000, never reused; homes made on the exporting
+  machine, else the controller; keys in /etc/llmgw/ssh-keys/%u through an sshd drop-in), SSH keys
+  from the portal's Profile or `ctl slurm users keys`. A user already on a machine with another UID
+  is left alone and said. The site has AD (not yet live): users from it, through sssd, are the next
+  step (`slurm.users` empty today leaves them to the site).
+- Logging in to a node, done 2026-11-14: llmgw's users reach the controller freely, another Slurm
+  machine only while a job of theirs runs there (pam_slurm_adopt, installed by the agent; PrologFlags=
+  Contain). Only the users listed in the agent's users file are held: root and the machine's own users
+  log in as before, so a wrong line never locks out who would fix it. pam_systemd would move the session
+  out of the job, so sshd's stack takes the session file in place of its include, with pam_systemd
+  skipped for the listed users alone; where that cannot be done safely (nested includes, jumps across
+  it) they are held to having a job, not to its cores and cards, and the agent says so. Tried on the
+  VMs: ayse without a job refused ("no active jobs on this node"), with one in
+  job_N/step_extern (nproc 1), cut off when it ended; mehmet refused; admin's session unchanged.
+  Left: sites whose users come from AD (sssd) are not held yet; a group of theirs would be the list.
+- A GPU node, tried 2026-11-14 on the PoC (cuda01, Quadro RTX 6000, Slurm 23.11.4): gres.conf
+  `Type=rtx6000 File=/dev/nvidia0`, slurm.conf `Gres=gpu:rtx6000:1`. `srun --gres=gpu:1`:
+  CUDA_VISIBLE_DEVICES=0, the one card; torch found it and ran 14.1 TFLOPS (fp32 matmul). Without
+  --gres: "No devices found", torch none (ConstrainDevices holds). `--gres=gpu:2`: refused. sacct has
+  `gres/gpu=1` in AllocTRES. Found: lending a node whose GPU model has nowhere else to go waits out
+  the drain grace (5 min) before the replica stops, and `ctl slurm status` says only "1 replica(s)
+  still to move off"; it should say it waits for a replacement that cannot be placed, and until when.
+  Done (c3a75dc): it says the time.
+- Jobs in containers, done 2026-11-14: `slurm.containers` (`ctl slurm containers on`). Each agent
+  installs enroot 4.2.1 (NVIDIA's .deb/.rpm, sha256 pinned) and builds pyxis 0.24.0 (source tarball,
+  sha256 pinned) against its own Slurm (gcc, make, libslurm-dev); on a machine with nvidia-smi,
+  nvidia-container-cli from NVIDIA's repository, its key's digest pinned. enroot.conf keeps each
+  user's runtime, cache and data under /run/enroot and /var/lib/enroot (no login session needed), the
+  shared path bind-mounted in every container. The leader writes plugstack.conf ("required") only
+  once every machine reports both, since a machine without the plugin would fail every job. Tried on
+  the VMs: on in ~30 s; ayse's srun --container-image=ubuntu:24.04 on n1 ran as uid 200001 with
+  /shared inside and nproc 1; a two-node sbatch ran a rank on each. Seen once: a node's pull timed out
+  reaching auth.docker.io and the job failed (network, not pyxis). Not tried: GPUs in a container
+  (cuda01's agent is an older build; needs the peer protocol), Red Hat.
+- AD end to end, tried 2026-11-14 (decided: one AD per install, the existing auth.ldap group mapping, no
+  tree from AD, no sssd unless a site asks): the VMs' auth.ldap against Samba DC VM 165 (ldaps by IP,
+  CA in the settings, ml-team -> team:Test/arge). ada signed in on the portal with her AD password: an
+  account, in Test/arge, Slurm user ada UID 200003; her key added in the portal's Profile, ssh to the
+  controller ~20 s later, sbatch ran on n1 as ada. Disabled in AD: suspended in 56 s, the Linux user
+  gone in 89 s, ssh refused, home kept. Enabled: back in 32 s, same UID, key kept. Taken out of
+  ml-team: out of the team in 11 s; put back: in again in 56 s.
+- P6, Slurm's jobs in llmgw's accounting (written 2026-11-12, not started). The tenancy tree to Slurm:
+  each project an account (`sacctmgr add account`, parent its team and organization), each member a user
+  of it, kept by the leader from the store; `sbatch --account=PROJECT`. The records back: the leader reads
+  `sacct --json` for jobs ended since the last read (through the controller's agent) into a
+  `batch_usage` table (job, user, project, start, end, GPU-seconds by class, CPU-seconds, memory), priced
+  per GPU-hour and core-hour into cost_micros, summed with the requests' usage on the usage pages and in
+  budgets. Budgets enforced: a project's remaining budget as GPU-minutes in its account's GrpTRESMins, so
+  Slurm holds its new jobs once spent. Needs the same users on every node (open: LDAP/sssd or
+  llmgw-made). Also found: slurm.conf
+  wrote every node State=DRAIN, so each slurmctld start (on any change of the files) drained a lent node
+  until its agent resumed it, up to a minute; a lent node is now written without it. Not tried yet: a GPU node (gres).
+- Failover while lent, tried 2026-11-11 (v0.16.3-23): the first try killed the job (NODE_FAIL). A new
+  leader had heard from no node yet and sent the controller a slurm.conf without them; slurmctld,
+  restarted on it, dropped node05 and its job. Fixed (141a1f4, 0774e90): slurm.conf's nodes and the
+  controller's address are kept in cluster_state (`slurm-nodes`) and a new leader starts from them; with
+  nothing kept (an upgraded cluster) or the store unread, the controller's files wait up to 2 min for the
+  nodes; a node is told nothing until its controller's address is known (no part stops its slurmd). Tried
+  again: a 5-minute job on lent node05, leader 140 stopped, 150 took over; slurmctld not restarted, the job
+  ran to its end; reclaimed during it, node05 was llmgw's within seconds of its end. Also 3387f97: srun in
+  SrunPortRange=60001-63000, and Slurm's ports in llmgw's firewall for the cluster's networks.
+- Built 2026-11-11: P1 (`internal/slurm`, `slurm.*` settings), P2
+  (`internal/agent/slurmnode`, proto Slurm/SlurmStatus), P3 (`internal/control/slurmsvc`, wired in the
+  leader; `GET /api/v1/slurm`), P4's commands (`ctl slurm status|setup|lend|reclaim|off`). Decisions on the
+  way: a node not heard from or with Slurm jobs is cordoned, not drained (nothing of llmgw's to move, and a
+  new leader must not move models off every Slurm node); jobs are counted from the node's own slurmstepd,
+  so a controller that is down neither hides them nor, with none, keeps the node from llmgw; a drain whose
+  reason does not start with "llmgw" is a Slurm administrator's and never resumed. Left: ~~the TUI screen~~
+  (done 2026-11-11: console **Slurm**, lend/take back on a row), P5,
+  a try on the PoC (needs agents built from this; slurm-wlm 23.11 on Ubuntu 24.04), GPU minor numbers are
+  taken to be nvidia-smi's index.
+- In the first install too (2026-11-11): `llmgw server init --slurm`, and a Slurm part on the choices
+  screen, write `slurm: {enabled, controller: <this hostname>, all_nodes: true}` into server.yaml.
+  `slurm.all_nodes` (also `ctl slurm setup --all-nodes`) runs slurmd on every node in the leader's
+  inventory, those that join later too; none is lent. Refused with `--no-node`, or a hostname Slurm does
+  not take.
 
 ## D. Facts worth knowing before starting
 
-- Admin API: `127.0.0.1:9090`, no auth, `ctlClient` in `internal/cli/ctl.go`.
+- Admin API: `127.0.0.1:9090`, behind the token in
+  `/var/lib/llmgw/admin-token` (A5), `ctlClient` in `internal/cli/ctl.go`.
   Metrics on `:9091` with `/prometheus/targets` beside them.
 - Store: `internal/store`, migrations under `internal/store/migrations/{sqlite,postgres}`;
   the PoC runs PostgreSQL under Patroni.

CHANGELOG.md

@@ -18,7 +18,9 @@ is only released once its **exit criterion** is met, not when its code lands.
 | v0.11.0 | — | High availability from setup, the setup and console screens, usage sums and per-tenant alerts | released |
 | v0.12.0 | — | The cluster's settings in the store, in force when changed; the firewall switch; wipe | released |
 | v0.13.0 | — | The whole system stopped and started from one command; Qwen3Guard judges, on a GPU and CPUs together | released |
-| v0.14.0 | — | Knowledge bases: documents from people's own machines, kept apart per tenant, answered from | released (v0.14.1: adapters, the tree over the API, the firewall on every machine; v0.14.2: pictures from ComfyUI; v0.14.3: 3D models as jobs) |
+| v0.14.0 | — | Knowledge bases: documents from people's own machines, kept apart per tenant, answered from | released (v0.14.1: adapters, the tree over the API, the firewall on every machine; v0.14.2: pictures from ComfyUI; v0.14.3: 3D models as jobs; v0.14.4: any ComfyUI workflow as a job, presets, licences; v0.14.5: picture and 3D models that add themselves; v0.14.6: setup that explains every part and lets it be chosen, the gateway on all three HA machines; v0.14.7: the judge sized for its machine; v0.14.8: sign-in with Active Directory, several keys for one outside provider; v0.14.9: sign-in set up from the console, a node's console that shows the machine) |
+| v0.15.0 | — | The company's MCP tools and VS Code through the gateway; guardrails that ask outside services and put redacted values back; answers to similar questions | released (v0.15.1: Strata, a 125B model of experts on one card; a second mirror kept in step; pickled weights refused; a served model's context changed; v0.15.2: a served model moved to more cards; v0.15.3: a model stopped and kept, or removed with its files) |
+| v0.16.0 | — | Weights pinned to a commit and checked against what Hugging Face published; documents embedded by the gateway; Strata offered where models are chosen; cluster upgrades that say how far they are | released (v0.16.1: the mirror's Strata tab, Strata's ready-made packs; v0.16.2: standbys that answer clients, FreeToken, a model fetched once per node; v0.16.3: replicas per profile, a served model's settings on the mirror page, the CPU's speed before serving) |
 | v1.0.0 | 5 | What phase 5 has left: precise KV routing proven on vLLM, the block cost picker, defragmentation across NVLink domains, prefill/decode disaggregation | planned |
 
 Until v1.0.0 the minor version is the breaking one: configuration formats and Go
@@ -28,6 +30,937 @@ APIs change between minors without a deprecation period.
 
 ### Added
 
+- **An organization's admin.** Besides member and manager, a membership of
+  an organization may now be `admin` (`llmgw users join … --org Acme --role
+  admin`, the role chosen on a membership on the people page, the console,
+  the admin API). An admin sees the organization's page as a manager does and
+  also changes the organization from it: adds, renames and removes its teams
+  and projects, adds people with an approved account by their address and
+  takes them out, makes its members managers or admins, and revokes keys spent
+  in its projects. Limits, approving and suspending accounts, the
+  organization's name and the rest of the cluster stay with the cluster's
+  administrators; an admin cannot take away their own role. Every change is in
+  the audit log.
+- **Slurm beside llmgw, for training and batch work.** `llmgw ctl slurm setup
+  --controller NODE --nodes NODE,...` (or `--all-nodes`, or `server init
+  --slurm`) has the agents install munge and Slurm from the machines' package
+  sources: slurmctld on one node, slurmd configless on the others, slurm.conf
+  and gres.conf written from what each node reports. Every node stays llmgw's,
+  drained in Slurm, until `ctl slurm lend NODE`: its replicas move off, then
+  Slurm runs jobs there. `ctl slurm reclaim NODE` takes it back once its jobs
+  end. `ctl slurm status` and the console's **Slurm** show each node. The
+  controller also keeps the jobs' records: slurmdbd, in a MariaDB on that
+  machine alone (Slurm takes MySQL or MariaDB only). Slurm's ports
+  (6817–6819, srun's 60001–63000) answer the cluster's networks only, as
+  llmgw's own do.
+- **Slurm's shared files and users.** `slurm.shared` (`ctl slurm shared`) is
+  one path on every Slurm machine: the site's NFS export mounted, an llmgw
+  node exporting it over NFS (what `server init --slurm` chooses, `/shared`
+  from the first machine), or a file system the site mounts. `slurm.users:
+  llmgw` (`ctl slurm users on`) makes each active account a Linux user on
+  every Slurm machine, the same login and UID (from 200000, never reused)
+  everywhere, its home on the shared path; people add SSH keys in the
+  portal's Profile (`ctl slurm users keys` for an operator). They log in to
+  the controller; to another Slurm machine only while a job of theirs runs
+  there, their session put in the job (pam_slurm_adopt), so nobody uses a
+  node's cards outside Slurm, nor on a node llmgw serves from. Root and the
+  machines' own users are not held to it.
+- **Slurm's jobs in containers.** `llmgw ctl slurm containers on`
+  (`slurm.containers`) installs enroot and pyxis on every Slurm machine, at
+  pinned versions checked against their digests, and nvidia-container-cli on
+  GPU machines; Slurm loads pyxis once every machine has it, and then `srun
+  --container-image=...` runs a job in that image, as its user, with the
+  shared path inside.
+
+### Changed
+
+- **Forms only for those who may send them.** A viewer sees the people,
+  limits, mirror and cluster pages without their forms, an operator the
+  mirror's and a node's engines but not the rest; before, they were drawn and
+  answered "not found".
+- **Smaller forms.** A label sits close to its field, a field is as wide as
+  what goes in it rather than the page, buttons are smaller, and the short
+  forms — a new key, a password, connecting an editor, a capacity or a limit —
+  sit on one line or two instead of a screen.
+
+### Fixed
+
+- **The serve forms start at the longest context that fits.** The mirror
+  page's form began every GGUF model at 4096 tokens, and the console's at
+  4096 or the plan's figure, whatever the model was trained for. Both now
+  start at what it was trained for, as long as its KV cache fits where the
+  form places it beside the weights, else at the most that does, in whole
+  thousands; the mirror page's llama.cpp slots are as many of that, up to
+  four, as the room left holds, and its memory figure is for that pool. A
+  plan says what the model was trained for (`trained`). Strata stays at
+  its 32768; vLLM's form is unchanged.
+- **A model that failed to start was started in the same place again.**
+  Placement given the same inputs chooses the same node and card, so an
+  on-demand model whose replica failed to load was retried there, past each
+  backoff, until it was parked, with another card or node free all along.
+  The control plane now remembers where a model's replicas failed (node and
+  cards, up to eight places) and places the next one elsewhere when anywhere
+  else fits; with nowhere else it is tried there as before. Coming up, a
+  change to its definition or `ctl retry` forgets the places. The log says
+  `placing a model away from where it failed to start`, with both places.
+- **Deleted model files came back after a control plane restart.** An agent
+  registers with the files its startup scan found, and sends the list again
+  only when it changes; a file deleted after the agent started was listed
+  again by a control plane that restarted, until the next change. A new
+  stream now gets the list on its first heartbeat.
+- **`ctl weights` offered models in use for deletion.** A model of several
+  files (a set, its address ending in `/`) is fetched into a directory named
+  from the address without the slash; the sweep named it from the address
+  with it, so every set in the catalog looked like nobody's. On the PoC that
+  was 330 GB of six catalog models, the qwen3.8-flash-next Q8_0 among them;
+  `--collect`, and `ctl model remove --purge` of a set, would have deleted
+  them. Both now compute the name the fetch does.
+- **A sign-in took an organization's manager role away.** A group from single
+  sign-on, SCIM or Active Directory that puts someone in an organization wrote
+  the membership again at every sign-in or sync, as a plain member: a manager
+  made one on the people page was one only until they next signed in. A
+  membership the directory names is now added if missing and otherwise left
+  as it is.
+- **The limits form on the people page's panel** was laid out on one line, its
+  title squeezed into a column.
+- **Deleting the model files nothing uses said what it was asked to free,
+  not what it freed.** The console's "Delete model files nothing uses" and
+  `ctl weights --collect` added up every file the nodes were asked to
+  delete as soon as the request was sent, and a model of several files —
+  fetched into a directory of its own — was refused by the node as "not
+  empty", so a sweep said tens of gigabytes freed and nothing went. The
+  node now deletes such a directory whole (a link only as a link), and the
+  sweep waits for the nodes to list their files again: freed is what they
+  no longer list, and what they kept is said beside, with where to read
+  why. Removing a model with its files reports the same way.
+
+## v0.16.3 — 2026-11-03
+
+Replicas counted per profile — seven on cards and four on CPUs under one name, the cards filled first — and every
+setting of a served model from the mirror page; how fast a model answers on the CPU, said before it is served, by
+the threads it is given rather than the whole machine's memory.
+
+**Upgrading.**
+- No migrations. `llmgw ctl upgrade v0.16.3`: the control planes carry the replicas per profile, the model card and
+  the speed in the serve form and in `ctl model plan`; the agents carry the measurement of one core's memory rate,
+  without which the speed is reckoned the old way, from the whole machine's.
+
+### Added
+
+- **Replicas counted per profile: seven on cards and four on CPUs, under one
+  name.** `llmgw ctl model replicas <model> gpu=7 cpu=4`, the console's
+  Replicas per profile… and the mirror page's model card set how many of a
+  model's replicas run on each of its profiles, rather than every replica
+  on the best profile that fits. For a GGUF model with no profile on CPUs
+  one is added (`--cpu-memory`), preferred below the cards, so the router
+  fills the cards first and sends to the CPUs once they are full. A change
+  starts the replicas short before stopping those over: the model never
+  runs fewer. `ctl models` shows the counts; a scaling of another total
+  ends them.
+- **Every setting of a served model from the mirror page.** A served
+  model's row has a Settings link, and its card now changes the context
+  and longest answer, the cards a replica takes, the replicas on each
+  profile (with the CPU's speed beside), stops and starts it, restarts its
+  replicas, and takes it out of the catalog with its files or without —
+  beside the replicas, class, owner, memory and adapters it held before.
+  Stopped on the row is now the model stopped and kept, started again as
+  it was, rather than its scaling set to nothing.
+- **How fast a model answers on the CPU, before it is served.** The serve
+  form says, as the cores are changed, about how many tokens a second one
+  reply gets on the CPU and on one card of the fleet's fastest class, and
+  warns when it is slow (under 15) or very slow (under 5, a reply of 500
+  tokens taking minutes). Very slow on the CPU is refused until "run on the
+  CPU anyway" is ticked, GPUs free or not. `llmgw ctl model plan` and the
+  console's Hugging Face screen show the same figure for every way a model
+  can run, and `ctl model serve --cpu` says it in its log.
+- **The speed follows the threads a replica is given.** Each agent now
+  measures what one core reads beside what all of them read: a replica of
+  four threads on a large server reads at about four cores' rate, not the
+  server's, which the estimate used to assume and overstated several times.
+  Calibrated on the PoC by llama-bench at 1, 2, 4 and 8 threads (within 15%),
+  and a model of experts reads its experts' share, from the GGUF header.
+  Agents need upgrading to report one core's rate; until then the old
+  estimate is used.
+
+## v0.16.2 — 2026-11-02
+
+Standby control planes that answer clients themselves, with rate limits and budgets kept across them; FreeToken, an
+engine for models larger than the card; a served model's longest answer changed in place; and fixes found putting
+the router under load and fetching large models: a multi-file model fetched once per node however many replicas
+start there, downloads that carry on where they were cut off, and a router that does twice the work on two cores.
+
+**Upgrading.**
+- No migrations. `llmgw ctl upgrade v0.16.2`: the control planes carry active-active routing (off until
+  `llmgw ctl config set router.active_active true`) and the router fixes; the agents carry the set fetched once
+  per node and the set's progress and checking; the mirror carries the resumed downloads.
+
+### Added
+
+- **Standbys that answer the client API themselves.** With
+  `router.active_active: true` in the cluster's settings, a standby control
+  plane routes `/v1` to the leader's replicas instead of passing every
+  request to the leader, so more machines carry the traffic. It hears the
+  replicas from the leader every two seconds over the agent port (no engine
+  keys travel) and reports back what it used, so the leader does not evict
+  a replica busy behind a standby, and the on-demand models it was asked
+  for. Rate limits are the cluster's: each router's request and token
+  counts reach the others through the leader every two seconds, so a key
+  spread over several control planes gets its limit, not that many times
+  it. Budgets are brought up to the store's every 30 seconds. Off by
+  default; the portal, `/admin` and `/mcp` still go to the leader.
+- **A served model's longest answer, changed in place.** `llmgw ctl model
+  context <model> --max-output N` (and `max_output` on `POST
+  /api/v1/models/{name}/context`, and the console's Context length… form)
+  sets the longest answer on its own: the catalog's limit, in force at once,
+  with no replica started again. With a context it is set with it. Strata's
+  context can now be changed too (its setup's `--context`), so neither
+  needs the model removed and added again.
+- **FreeToken, an engine for models larger than the card.** Mixture-of-experts
+  models several times a card's size, from one card, with the experts in RAM
+  and streamed over PCIe — read from Hugging Face safetensors checkpoints as
+  published (DeepSeek V4 Flash, GLM-5.x, Qwen3.6/3.8, gpt-oss, Gemma 4,
+  MiniMax), with batching. Installed like vLLM on every node it runs on: a
+  card of compute capability 8.0 or newer, driver 580+, x86-64, and the CUDA 13
+  toolkit it compiles kernels with; older cards are passed over (tried on a
+  Turing card, where 0.2.1 has no kernels). Profiles are one card
+  (`engine: freetoken`); the key goes in its environment, and flags that read
+  files or fetch are refused. Readiness now also waits for an engine whose
+  `/health` answers 200 while saying `loading` in its body. FreeToken is
+  followed release by release: a weekly workflow fails when PyPI has a version
+  newer than `DefaultEngines` installs.
+- **How much of each GPU works, beside whether anything does.** The busy
+  figure is nvidia-smi's `utilization.gpu` — the share of time any kernel
+  ran, which a kernel on one SM fills — and it is now said to be that: on
+  the dashboard, the node page, the metric's help and Grafana. Beside it,
+  each card's draw against its power limit (the efficiency table, the fleet
+  and node charts, the console's GPU work chart), and where the card's
+  profiling counters can be read, its SMs, SM occupancy, tensor cores and
+  memory interface: `llmgw_gpu_sm_active_percent`,
+  `llmgw_gpu_sm_occupancy_percent`, `llmgw_gpu_tensor_active_percent`,
+  `llmgw_gpu_dram_active_percent`. The agent reads them from NVML's GPM on
+  Hopper and later, and from a dcgm-exporter on the node elsewhere
+  (`discovery.dcgm_exporter`, default `http://127.0.0.1:9400/metrics`, `off`
+  for none). The GPU map colours by this "work" by default — SMs active
+  where read, power against the limit otherwise — and keeps busy as a mode.
+  On cuda01 decode read 92% busy with the SMs at work 69% of the time.
+
+### Fixed
+
+- **Several replicas of a multi-file model on one node fetched it once
+  each, over one another.** Four replicas placed on a node's four cards
+  wanted the same safetensors set at the same moment, each wrote the same
+  `.part` files, and all but one failed to start ("rename … no such file")
+  until a retry found the set finished. A set is now fetched once per node
+  and every replica waiting on it gets it, as single files always were.
+- **A model of several files showed one file's download as the whole.** A node fetching a set reported the part it
+  was on — "59% of 46.1 GB" for a 188 GB model — instead of the set: it now reports the set's bytes against its
+  whole size.
+- **A set's files checked with nothing said, one at a time.** After the last byte arrived a node checked each file
+  against the mirror's checksum, 195 GB on one core for eight minutes, while the replica still read "fetching 100%".
+  It now checks four at once and says so: `checking 40% of 195 GB` in `ctl replicas`, "Checking the weights" in the
+  chat page; and a check that takes long is not taken for a stuck download.
+- **A mirror download the registry cut off started again from nothing, or not at all.** Hugging Face's CDN ended a
+  60 GB transfer at 23 GB (`stream error … CANCEL`), the mirror deleted what had arrived, and `ctl model serve`
+  failed. The mirror now tries again up to six times, carrying on from where it got to (`Range`), keeps what arrived
+  when it gives up, and carries on from it after a restart too.
+- **A model from an outside provider took three or four requests at once.** Admission read every request in flight to
+  an endpoint it does not measure as a queue, so an external model (OpenAI, another llmgw cluster) was held to what a
+  saturation of 0.85 allows of "in flight / 4"; the rest waited and timed out with 504. An outside provider queues for
+  itself and says 429 when full, which its breaker already hears: it no longer counts towards saturation.
+- **A valid key told to slow down under load, every thirty seconds.** When a busy key's cached entry expired, every
+  request in flight missed the cache together and each took one of the address's allowance for unknown keys: past a
+  hundred or so streams at once, a valid key got 429 "too many unrecognised api keys". Requests missing the same key
+  at once now share one read and one allowance.
+- **A streamed answer was one write to the client a token.** Each event was flushed alone, a system call each: a
+  third of a busy router's CPU. Events already received from the engine now go out in one write, with no token
+  waiting for another; on the PoC's two-core leader the ceiling rose from ~575 to ~700 streams' requests a second,
+  and the first token at 128 at once from 91 to 45 ms.
+
+## v0.16.1 — 2026-10-30
+
+The mirror page's Strata tab, Strata's ready-made packs served with Strata fetching them, and two fixes found
+serving Strata's models: repositories that keep each size in a directory, and DeepSeek on cards its engine does not
+run on.
+
+**Upgrading.**
+- No migrations. The control planes carry the tab, the plan and the serving; the agents carry the refusal of a
+  Strata profile its card cannot run (`llmgw ctl upgrade v0.16.1`). Mirrors need nothing new.
+
+### Added
+
+- **The mirror page's Strata tab.** Every family and size Strata runs, with its download, the RAM it keeps and where
+  it would run here; a GGUF size fetched into the mirror in one press (checked against Hugging Face's checksums),
+  and the packs Strata fetches itself — NVFP4, DeepSeek V4.1 — served at once, their repository the source.
+  `ctl model plan` and `ctl model serve` take those repositories too, and the explorer names a `strata-ple`
+  repository for what it is: part of a pack Strata knows, or one Strata 0.2.7 does not.
+
+### Fixed
+
+- **A repository that keeps each quantisation in a directory of its own** (ISTA-DASLab's `IQ2_XS/`, `IQ3_S/`) listed
+  no weights: its file tree is now read recursively, by the explorer, the plan and `ctl model serve`.
+- **DeepSeek V4.1 offered on a card its engine was not built for.** Strata's DeepSeek engine runs on RTX 30 (compute
+  8.6) and newer; on a Turing card the tab and the plan offered it, and Strata's setup refused it once placed. Now
+  neither offers it there, the tab says why, and the agent refuses such a profile before starting it.
+
+## v0.16.0 — 2026-10-30
+
+Weights pinned to the commit they were added at and checked, by the nodes and the mirror, against the sha256
+Hugging Face published; the documents `llmgw-kb` sends embedded by the gateway; Strata offered by the plan, `ctl
+model serve`, the explorer and the serve forms; and cluster upgrades that tell the nodes together, say how far they
+are, and do not fail a node that came back on the version.
+
+**Upgrading.**
+- No migrations. Upgrade the control planes, then the nodes (`llmgw ctl upgrade v0.16.0`), then each mirror's host
+  (`sudo llmgw upgrade v0.16.0 --service llmgw-mirror`).
+- A node checks what it fetches against the catalog's checksums only once it runs this version; until then it
+  fetches as before.
+- A model added from Hugging Face is now pinned when it is added, which needs the registry reachable from the
+  control plane; one it cannot be looked up in is refused, saying why. Models already in the catalog are left as
+  they are.
+- A Strata profile must say `--size` as well as `--family`, and both must be ones Strata runs.
+- `llmgw-kb` sends text alone to a gateway of this version; an older `llmgw-kb` keeps working, sending vectors.
+
+### Added
+
+- **Strata offered where models are chosen.** For a repository of a family Strata runs, `ctl model plan` offers "one
+  card with Strata, experts in RAM" with the RAM it takes, `ctl model serve --engine strata` (or the plan's choice)
+  builds its profile, the Hugging Face explorer lists it as running here, and the mirror's serve form serves such a
+  version on Strata. Strata's families and sizes are known: the catalog refuses one it does not run and reserves the
+  RAM a size keeps its experts in.
+- **The gateway embeds the documents `llmgw-kb` sends.** A document may be sent as its passages' text alone; the
+  gateway embeds them through its own `/v1/embeddings` as the caller, 32 at a time and four batches at once,
+  waiting out a busy engine rather than failing the document. About a kilobyte a passage leaves the machine instead
+  of a vector down and up again, some thirty times more. `llmgw-kb` does so with a gateway that says it can
+  (`server_embeds` in a collection's answer); `--embed-here` keeps the old way.
+- **Weights pinned to a commit, and checked against what Hugging Face published.** A model added from Hugging Face
+  (`hf:`, a resolve URL on a branch, or a repository vLLM or SGLang downloads) is pinned to the commit it is at
+  then — the file's URL at that commit, or `--revision` — and each file's published sha256 goes in its origin. A node
+  refuses and deletes a copy that is not it, naming both checksums; the mirror does the same for what `ctl model
+  serve` fetches, and `llmgw mirror add` for a Hugging Face URL (`--sha256` for others). Nodes need this version to
+  check; older ones fetch as before.
+
+### Fixed
+
+- **A cluster upgrade that sat silent, or stopped on a node that had upgraded.**
+  - The nodes of a wave are told together, each saying when it is done, and what is still awaited is said every 30
+    seconds; before, each was told in turn, up to five minutes each, with nothing on the screen.
+  - The control plane takes the node registering again on the version as done: the agent's answer, sent a moment
+    before it exits, could be lost, and a node that had upgraded was failed as "older than v0.10.4".
+  - The agent's download no longer ends with its connection (a control plane restarting, leadership moving), and the
+    control plane waits longer than the download may take; a second request while one runs is refused.
+  - A cordoned or draining node upgraded is done; a node whose idle model stopped meanwhile is done once nothing on it
+    is starting; a node not connected is passed over and named at the end rather than stopping every node after it.
+  - The console's node upgrade waits as long as the control plane, and offers no development version.
+  - systemd never gives up restarting the agent (`StartLimitIntervalSec=0`), in units written from now on.
+
+- **Strata.** One named by hand in `agent.yaml` was never reported, so never placed on; its engine, which leaves the
+  server's process group, is now stopped with it; Pascal and Volta, which its CUDA 12 build runs, were refused.
+- **A name with a slash in it, asked through a standby control plane.** `llmgw ctl kb delete Demo/hr` on a standby
+  was answered 405: the standby passed the request on to the leader with `Demo%2Fhr` unescaped, two segments instead
+  of one. The path now reaches the leader as it was sent.
+
+## v0.15.3 — 2026-10-29
+
+A model stopped and kept, or removed with its files; the second mirror fixed where it was tried.
+
+**Upgrading.**
+- No migrations. The control planes carry stop, start and purge; the nodes need nothing new.
+- The mirrors carry the fixes: on each mirror's machine, `sudo llmgw upgrade v0.15.3 --service llmgw-mirror`.
+  A mirror with a peer then works out the checksums its older files lack, once, which reads every such file.
+
+### Added
+
+- **A model stopped and kept, and a model removed with its files.** `llmgw ctl model stop|start MODEL`: none of its
+  replicas run, it stays in the catalog with its files, and start runs it again scaled as it was; requests are told an
+  administrator stopped it. `llmgw ctl model remove MODEL --purge` (`DELETE /api/v1/models/{name}?purge=true`) also
+  deletes every version of it on the mirror, a second mirror following, and the nodes' copies once its replicas have
+  stopped; what another model uses stays. Both are in the console's model menu.
+
+### Fixed
+
+- **`llmgw mirror add-peer` on a mirror with a machine of its own.** It set the second mirror up and then
+  failed for want of the control plane's server.yaml; now it prints the `llmgw ctl config set mirror.peers`
+  to run on a control plane.
+- **Models fetched before checksums were kept reach the second mirror.** A mirror kept in step copies only what it
+  can check, and those files had no sha256; a mirror with peers now works theirs out and records them, once.
+- **A file removed with `llmgw mirror remove` is gone from the mirror's answers too.** The running mirror kept
+  reporting a file it had fetched as present for a week after it was removed, so a control plane could place a
+  model on a file that was not there; and the removal was not recorded, so a mirror kept in step copied it back.
+
+## v0.15.2 — 2026-10-29
+
+A served model moved to more cards, or fewer, while it is served.
+
+**Upgrading.**
+- No migrations, and nothing changes until the command is used. The control
+  planes carry it; the nodes need nothing new.
+
+### Added
+
+- **A served model moved to more cards, or fewer.** `llmgw ctl model gpus
+  MODEL N [--profile P] [--in-place]`, `POST /api/v1/models/{name}/gpus`, and
+  GPUs… in the console's model menu. How many cards of one node each replica
+  uses: tensor parallel for vLLM, SGLang and TensorRT-LLM, checked against
+  the model's KV heads, and the layers split for llama.cpp. Refused for a
+  count no node has, or whose cards would not hold the model and its
+  context, with the figures; its replicas are then started again on it, one
+  by one, beside the old — or, with `--in-place` when there is no room
+  beside them, each stopped first.
+
+## v0.15.1 — 2026-10-29
+
+A 125B model on one card, a second mirror, and weights that run code
+refused. Strata runs as an engine: a mixture-of-experts model with its
+experts in RAM, served from one 24 GB card. A second model mirror keeps in
+step with the first, so an HA install has no single machine left. Pickled
+weights are refused unless a model says otherwise. A served model's context
+can be changed in place. The explorer says how fast a model would answer
+here before it is fetched, and a card in the wrong PCIe slot is said so.
+
+**Upgrading.**
+- No migrations. Upgrade the control planes before the nodes, for the
+  pickle refusal below.
+- A model whose weights are in a pickle format (`.bin`, `.pt`, `.pth`,
+  `.ckpt`, `.pkl`) stops being placed: its node refuses the source. Give it
+  its safetensors or GGUF, or add it again with `allow_pickle: true`
+  (`llmgw ctl model add -f`) if its files were checked. Look at the
+  models' sources before upgrading the nodes: one ending in
+  `.ckpt` or `.bin`, or a mirror version holding `pytorch_model.bin`. Upgrade the control planes first: a node on
+  this version refuses those weights until its control plane can say
+  `allow_pickle`.
+- Strata is opt-in: nothing installs it until `engines` or an
+  `engine_overrides` entry names a version (`0.2.7`).
+
+### Added
+
+- **Strata as an engine: a 125B model of experts on one card.** Strata
+  (github.com/Niko1221/Strata) keeps a model's experts in RAM and the ones
+  used most in VRAM. On an RTX 6000 24 GB with 251 GB of RAM it serves
+  Qwen3.8-Flash-Next 125B (IQ2_XS) at 70–87 tokens a second to one client;
+  more clients share about that, its slots buying fairness rather than
+  throughput on one card. It runs a few families only — Qwen3.8-Flash-Next
+  and its variants, Qwen3.6-35B-A3B, DeepSeek V4.1-Flash — each named in a
+  profile's args with `--family` and `--size`
+  (`--family qwen --size IQ2_XS --context 32768`). The source is the
+  directory of the model's GGUF files, or the repository, which Strata then
+  downloads itself, pinned. The agent installs it on an NVIDIA node of
+  x86-64: its source pinned by digest, in a Python environment, and its
+  release's engine checked against the digest GitHub publishes — the CUDA 13
+  build for Turing and newer on driver 580, the CUDA 12.9 build for Pascal
+  to Ada, a V100 included, on driver 575. A replica prepares its model the
+  first time (minutes; its draft layer is a 5 GB download) and loads in
+  about a minute and a half after. One card of one node per replica, for
+  now; not in an offline bundle, its packages coming from PyPI.
+
+- **A served model's context, changed.** `llmgw ctl model context MODEL
+  TOKENS [--slots N]`, `POST /api/v1/models/{name}/context`, and Context
+  length… in the console's model menu. The catalog, the engine's arguments and
+  the memory change together, checked against what the model was trained for
+  and what its cards hold; its replicas are then started again one by one. The
+  models API shows `context_length` and `max_output`.
+
+- **A second model mirror, kept in step.** In a three-machine install the setup screens put one on a database machine
+  (`--second-mirror 1|2`). An install with one mirror gets one with `sudo llmgw mirror add-peer user@host`.
+  - **In step:** each mirror reads the other every thirty seconds. It copies what it lacks, checksum checked, and
+    removes what the other removed, even if it was down at the time (`llmgw mirror serve --peer`).
+  - **Control planes:** they name both (`mirror.peers`) and use whichever answers.
+  - **Nodes:** while one mirror is down, a node is given a model's file on the other.
+
+  This takes away the last single point of an HA install: with the first machine off, models are still added and
+  fetched.
+
+### Fixed
+
+- **An engine that computes in a process of its own group was not its
+  replica's.** A GPU replica read as running on the CPU, and the memory it
+  held on its card as somebody else's, which placement then kept free of
+  other work. A process the replica started, at any remove, is now the
+  replica's. And what a replica computes on, once sent, was never sent
+  again: a wrong reading kept from before an agent's restart stayed wrong.
+  It is sent again when it changes, and read again when a GPU replica was
+  kept as the CPU.
+
+- **VS Code's sign-in lost its place through the sign-in page**, so the
+  prompt to confirm came every time; a link to the same gateway by plain
+  http could take an https setting down to http; and a new build of the
+  extension was not installed over an old one (the command now says
+  `--force`).
+- **The install guide's recordings were not checked for secrets**, only
+  the page; every frame is now.
+- **MCP server fields could not be emptied**, an address could carry a
+  password, the cluster's own server could not be changed from the
+  console, and a grant to a key that does not exist was said done.
+- **A guardrail service two policies named was asked as the first said**,
+  not the stricter; a `null` answer read as safe; and with `restore` the
+  request log kept the answer with the redacted values back in it.
+- **What an engine downloads by a Hugging Face id was not held to the
+  pickle refusal** — an adapter by its id, a TensorRT-LLM model's; a model
+  with either now needs `allow_pickle`.
+- **`server init` run again on a machine already set up was refused its
+  memory**, as on an 8 GB install fixing its certificate with `--host`, and
+  judges that would not be installed (no mirror) were counted.
+- **Removing an MCP server left its sealed credential in the store**, and
+  replacing a credential under another header kept the old one sent. A
+  server's secrets now go when it no longer reads them.
+- **A cloned repository could point the VS Code extension at another
+  gateway.** `llmgw.gatewayUrl` is a machine setting now, never taken from a
+  workspace.
+- **A node's memory bandwidth could be measured in its cache** on a busy
+  node, and was 0 after a reconnect until the next heartbeat. The buffer is
+  at least eight times the processor's last cache.
+- **A project kept in the cluster could have its questions embedded outside
+  it.** The similar-question cache now asks the embedding model as the
+  question's project would, external fallbacks refused, and skips a model
+  that is itself outside. Its comparing no longer holds the cache's lock,
+  which every project's lookups waited on.
+- **Restored values were left as placeholders in streamed tool calls**, in
+  completions and in reasoning, and a value with a quote in a whole answer's
+  tool-call arguments broke their JSON.
+- **An outside guardrail service read what `redact` had taken out**, and a
+  streamed answer was sent to it twice. The docs now say a service stops a
+  streamed answer only under `stream: buffer_all`.
+- **A replica found slow once could stay unused.** Its time to first token
+  is forgotten after five minutes without an answer, so it is tried again.
+- **Weights that run code got past the refusal in four ways.** A mirror
+  version already on a node was used without being looked at again; a
+  `?query` or `#fragment` on an `hf:` source hid the file's name; `.pt2` was
+  not on the list; and a profile could write `--load_format`,
+  `--trust_remote_code` or an abbreviation such as `--trust-remote`, which
+  vLLM and argparse read as the flag, or give SGLang `--load-format`. All
+  refused now. TripoSR's preset says `allow_pickle`, without which it could
+  not be added, and a shared ComfyUI environment takes its first member's.
+- **A temperature written as a string was refused.** Since the GenAI traces
+  read it, `"temperature": "0.7"` got a 400 before reaching the engine; it
+  is read loosely again, for the trace only.
+- **A reused answer could carry another caller's personal data.** With
+  `restore: true` and the answer cache on, two callers' requests that
+  redacted to the same text shared one answer — with the first caller's
+  address put back in it. Such requests are now never answered from the
+  cache nor kept in it.
+- **Granting a server's tools could take others away; revoking could close
+  a level to every tool.** A level with no tool list sees what the levels
+  above grant; giving it its first list made it see only that list. And a
+  level left with an empty list was closed to every tool, which the screens
+  did not show. Now a first list starts from the nearest list above, taking
+  a server away leaves the rest or goes back to no list, the screens say
+  when a level above keeps granted tools out of reach and which levels are
+  closed, and the console reads and writes the level at once rather than
+  from what it loaded.
+- **Similar questions about different pictures got each other's answer.**
+  A question with a picture, a sound or a file in it is reused only when it
+  is the same request.
+
+### Added
+
+- **A card in the wrong PCIe slot is said so.** Each agent reads how every
+  card is linked — the fastest generation card and slot allow together, and
+  the lanes in use — against the card's own most, from the kernel (the
+  driver's figure is what card and slot allow together); a card on fewer lanes or
+  an older generation is a node warning, with what it carries and what it
+  could, on the dashboard, the console and in `llmgw doctor`.
+- **And on a web page, `/admin/mcp`.** For administrators: the same list,
+  adding a server with its credential sealed, testing it and listing its
+  tools, granting them in the tree beside that level's other grants and
+  limits, revoking, changing, replacing the credential, turning it off,
+  removing; and the line that points an agent at this gateway's `/mcp`.
+  In English and Turkish; every change in the audit log, the secret not.
+- **The company's MCP servers in the console.** A screen of its own: each
+  server with its state, who has its tools, its rate, the names of the
+  credentials it is sent and its address; add one with a bearer token or
+  a header kept sealed, test it and list its tools, grant them to a team,
+  project, organization, person or key beside that level's other grants,
+  revoke, change, replace the credential, turn off, remove.
+- **And on the cards, as this cluster would lay it out.** The explorer's
+  **GPU tok/s** and a repository's files give one reply's speed for each
+  layout the planner offers here — one card, tensor parallel across a
+  node's cards, llama.cpp's layer split, a pipeline across nodes on a
+  fabric — the fastest of them, with the layout in the tooltip. From each
+  card's memory bandwidth by its model name, NVLink or PCIe between a
+  node's cards, and InfiniBand or RoCE between nodes; a cluster of mixed
+  classes is judged class by class. A model of experts reads 1.6 times its
+  active share, as Qwen3-Coder-30B-A3B did on cuda01's RTX 6000 (144
+  tok/s; qwen3-vl-8b Q8: 62).
+- **How fast a model would answer on a CPU, before it is fetched.** Each
+  agent measures how fast its node's memory reads when it starts (`Host`
+  gets `mem_bandwidth_gbs`, the node's view `mem_bandwidth_gbs`); the
+  explorer's **CPU tok/s** column and a repository's GGUF files say about
+  how many tokens a second one reply gets on the fastest node that holds
+  the model — the bandwidth over the bytes a token reads, its active part
+  for a model of experts. Calibrated on the PoC against llama.cpp's own
+  timings: within 15% from 1.5B to 14B.
+- **The explorer lists what this cluster can run.** Judged by the nodes
+  as they are now — every card's memory and compute capability, every
+  node's RAM — through the same planner as serving a model, rather than a
+  size picked from a list. A model too big for any node, in FP8 without a
+  card of 8.9, in MLX, or in neither GGUF nor safetensors, is left out and
+  counted; the page says what it judged by. With cards, safetensors repositories are
+  listed beside GGUF. On a repository's page a file too big for the
+  cluster is shown with what it needs and no Fetch button.
+- **Weights that run code are refused.** Python pickles — `.bin`, `.pt`,
+  `.pth`, `.ckpt` — run whatever they hold when loaded, which is how the
+  malicious models on Hugging Face work. The node refuses such a source,
+  or a mirror version with one in it, before fetching it; adding such a
+  model is refused; vLLM and SGLang downloading a model by its id read its
+  safetensors only; the mirror page marks the file. `allow_pickle: true`
+  on a model lets it through on purpose. The mirror's explorer no longer
+  lists a repository with nothing a node loads (no GGUF, no safetensors)
+  while "runs here" is ticked, and a repository's pickle files are not
+  offered, only counted. BACKLOG M lists what is next:
+  checksums against Hugging Face everywhere, its own scan, trusted
+  publishers, engines without the internet.
+
+## v0.15.0 — 2026-10-25
+
+The company's tools and editors through the gateway, and guardrails that
+can lean on others. Its own MCP servers sit behind llmgw's `/mcp` under the
+guardrails, a rate and a breaker. VS Code gets the gateway's models from an
+extension the gateway serves. A guardrail policy can ask an outside
+service, and can put redacted values back in the answer. Chat goes to the
+replica answering soonest. A similar question can get the answer already
+given. A cached prompt can be priced apart. Traces follow the OpenTelemetry
+GenAI conventions, and Kong is written down from trying it. Also an install
+guide with every screen, at `/install`.
+
+**Upgrading.**
+- Migrations 0049 (`mcp_servers`), 0050 (their rate and breaker) and 0051
+  (`projects.cache_similar`), applied when the control plane starts.
+- `llmgw server init` now refuses a machine that would serve models without
+  the memory for them: the starter model and the judges need 12 GB on one
+  machine, where 8 GB was said before. `--no-model`, `--no-guard-model` or
+  `--no-node` make it fit. A running cluster is not checked again.
+- Traces carry the GenAI conventions' `gen_ai.*` attributes beside the
+  `llmgw.*` ones, which stay; docs/install.md, "Tracing", lists them.
+- Nothing else changes until it is turned on: `services` and `restore` in a
+  guardrail policy, `keys cache --similar` with
+  `router.cache.embedding_model`, `cached_per_1k` in a price.
+
+### Added
+
+- **Answers to similar questions.** `llmgw keys cache --similar 0.95` also
+  gives a chat question the answer already given to one like it in the same
+  conversation, compared by `router.cache.embedding_model`, as
+  `x-llmgw-cache: similar` with the similarity. Held in memory beside the
+  exact answers; nothing written to the store but the setting.
+- **A cached prompt costs the cached rate.** `cached_per_1k` in a price
+  charges the prompt tokens an engine or provider read from its cache (the
+  answer's `cached_tokens`) at their own rate; without it they cost what any
+  prompt token does, as before.
+- **Chat goes to the replica answering soonest.** The router keeps each
+  replica's time to first token, smoothed over its last answers, and the
+  `chat` profile scores by it beside the queue and the KV cache: of two
+  replicas of one model, the one on the slower card or the busier host gets
+  less, and a replica not yet measured is tried.
+- **llmgw and Kong, written down** ([docs/kong.md](docs/kong.md)): Kong 3.9
+  in front of llmgw by a plain route or its `ai-proxy`, each team's llmgw
+  key passed on; Kong as a provider behind llmgw; and that Kong 3.9 counts
+  no completion tokens of a streamed answer, so bill from llmgw's usage.
+- **A guardrail that is someone else's service.** A policy's `services`
+  asks Azure Content Safety, Lakera, NeMo or a company's own classifier over
+  HTTP: a body template, keys from files, dotted paths to where the answer
+  says unsafe and why, on input, output or both, block or flag, and the
+  policy's `fail` when it does not answer.
+- **Redacted values put back in the answer.** `restore: true` on a guardrail
+  policy's input: the model reads `[EMAIL_1]`, the client reads its own
+  address in the reply, streamed or not. Only the request's own values, held
+  in memory for its life; the output rules read the answer as the client
+  will.
+- **Traces in the OpenTelemetry GenAI conventions.** An inference request
+  and each attempt at an engine carry `gen_ai.operation.name`, the requested
+  and the answering model, `max_tokens`, `temperature` and `top_p` when set,
+  the session as `gen_ai.conversation.id`, and the input, output and cached
+  tokens, and an attempt the engine that answered as `gen_ai.provider.name`.
+  Tools that read the conventions show llmgw's calls as model calls with no
+  dashboard of their own. No prompt or answer text is ever in a span.
+- **llmgw for VS Code.** An extension of llmgw's own, served by every gateway
+  at `/setup/vscode/llmgw.vsix` with that gateway's address in it, made from
+  the binary itself: plain JavaScript, no dependencies, no build, and so no
+  marketplace needed at a site with no internet. It offers the gateway's chat
+  models to VS Code's language-model API — the chat's model picker, agent
+  mode, and every extension that asks VS Code for a model — with the sizes
+  and the tool and vision capabilities the catalog says, streamed, with tool
+  calls and their results both ways. **VS Code** is a tool on the account
+  page's **Connect your editor**: a key, the one-line install, and an **Open
+  in VS Code** link that hands the key to the extension, which keeps it in
+  VS Code's secret storage. **llmgw: Sign In** opens that page with a state
+  of its own; a link without it is confirmed first, the gateway named.
+- **An install guide with every screen, at `/install`.** From three bare
+  Ubuntu machines to a first answer: what each machine needs and how to
+  check it, the release fetched and checked, the setup screens one by one,
+  the system's state, and a first request with a key from the account page.
+  The screens are a real three-machine install, recorded and kept as
+  terminal frames — sharp at any size and a few kilobytes — with each
+  session played back on request; the administrator's password and every
+  key were masked before it was kept. `/about` says more of what llmgw does
+  (documents, guardrails, sign-in, high availability, pictures and 3D), and
+  its install steps are the ones of today, leading to the guide.
+- **The company's MCP tools under the guardrails, a rate and a breaker.**
+  - **The answer:** the caller's policies read a tool's answer as the prompt it becomes. An instruction planted in a
+    ticket is withheld as a pasted one would be, and personal data is redacted.
+  - **The arguments:** they are read before they leave, so a secret pasted into one does not reach the company system.
+  - **Rate:** `ctl mcp add --rate N` allows each caller N calls a minute to a server.
+  - **Breaker:** five failures in a row take a server out for a minute, its tools off the list.
+
+- **How fast the judge is, measured.** `llmgw ctl guard judge check` (and `POST /api/v1/guard/judges/check`) works
+  out whether the judge keeps up with requests. It times three judgments of a 280-token prompt, then a burst at once,
+  and sets both against the policies' wait. It says ok, tight or slow, about how many requests a second can be judged,
+  and what to do: more cores, more judges, a GPU judge, a longer wait, or `fail: open`. `server init` runs it once the
+  judge is up. If the machine is too slow it says so before the next steps, not after the first refused request.
+
+- **The company's own MCP servers through llmgw's `/mcp`.** `llmgw ctl mcp add jira --url … --secret-file
+  Authorization=FILE` registers one; `/mcp` then lists its tools as `jira.search` beside llmgw's own. Only keys granted
+  them see them: `llmgw keys limits --tools 'jira.*'`, patterns at any level of the tenancy tree. None granted anywhere
+  is none. The server's credentials stay on the control planes, kept in the store sealed or read by reference. The
+  server is told the calling key, person and project. Each call is bounded in time and size, and its progress passes
+  on to a streaming client. Every call is in the audit log with its arguments' size and digest, never their text. The
+  admin API has `/api/v1/mcp/servers`.
+- **The cluster as a read-only MCP server.** Registered as `llmgw ctl mcp add llmgw --url llmgw:admin` and granted
+  as `llmgw.*`, its tools say what `llmgw ctl` does:
+  - `cluster_status`, `list_nodes`, `list_models`, `list_replicas`, `explain_placement`, `node_logs`;
+  - nothing that changes the cluster.
+
+  It is answered in the control plane's own process, so there is no token to keep and every control plane serves it.
+  The admin API serves the same at `POST /api/v1/mcp`. `ctl mcp add --bearer-file` keeps a token, sent as
+  `Authorization: Bearer`.
+
+- **Large models across cards and nodes.** For a model of experts on more than one card, the plan now adds vLLM's
+  `--enable-expert-parallel`. It reads the expert count from `config.json` under any of its names, and sizes a model
+  that also reads pictures by its `text_config`. Several nodes are offered in any number, not only powers of two. Each
+  way now says what it needs of a card, part by part (weights, context, engine), and of all its cards together. This
+  shows in `ctl model plan`, in the console's Serve a model, and in a serve job's first step. The serve form checks that
+  a longer context still fits. `ctl model serve --env KEY=VALUE` puts engine variables such as `NCCL_DEBUG=INFO` on
+  every node. docs/large-models.md has the rules, the commands, and a checklist for bringing a model up on a fabric.
+
+- **Machines of the judges' own in a three-machine install.** The setup screen asks for up to two judge machines after
+  the store's two; `server init` takes `--judge-node user@host` for each. Each is added as a node labelled
+  `llmgw.role: judge`, and the judge's profile selects it, so the judges run there alone. Under load on a 4-core
+  control plane that was also a node, 38 of 60 requests were refused for want of a judgment. With the judges on a node
+  of their own, 2 were. A node with an `llmgw.role` label takes only profiles whose node selector asks for that role,
+  so no other model is placed on the judges' machine.
+
+### Fixed
+
+- **A machine too small for the models it would serve is said so first.**
+  The control plane that is also a node, serving the starter model and two
+  judges, needs about 10.5 GB. On 8 GB the install finished every step and
+  then had no room for the starter. Now the setup's checks fail its memory
+  with what it would serve, and `server init` refuses before it installs
+  anything, naming `--no-model`, `--no-guard-model` and `--no-node`.
+
+- **Turkish pages are Turkish to the end.**
+  - **Errors:** those from the store, catalog and mirror that a page shows ("not found", "not empty", "already in the
+    catalog", the settings changed meanwhile, and others) are now in the page's language.
+  - **Percent:** a percentage is written the language's way: `%93` in Turkish, `93%` in English.
+  - **Footer:** the time at the foot of a page is in the language's date form.
+  - **Stale forms:** they are refused with a sentence in that language rather than "bad request".
+
+- **`llmgw keys limits` no longer clears the limits it was not given.** Setting one limit at a scope wrote the whole
+  row, so `--tools 'jira.*'` alone took away that scope's rate, budget and model list. Only what is given changes now.
+  `--replace` writes every limit, as the console's limits form does.
+
+- **The plan refused tensor parallelism past a model's KV heads.** Qwen3-235B, with four KV heads, had no way to run
+  on eight cards, though vLLM gives each card a copy of a head. A width that is a multiple of the KV heads is now
+  offered, as is one that divides them.
+- **A big model on a slow mirror was failed while it downloaded.** The startup timeout counted the download, so
+  hundreds of gigabytes at a few hundred MB/s ran past twenty minutes and the replica was started again from nothing.
+  The timeout now runs from when a node's download last moved.
+
+- **The judge init installs is no longer replaced by the leader's own.** Just after the control plane started, the
+  leader saw a policy naming a judge that init was still fetching, and served one itself. Its entry replaced init's, so
+  a second judge ran with the leader's sizing, and with judge machines it ran off them. The leader now waits ten minutes
+  after it starts before it serves a judge itself. If someone adds one while it fetches, theirs stays.
+- **A judge the leader serves by itself is sized as init's is.** When a guard policy names a judge the catalog lacks,
+  the leader serves it. That judge reserved 3 GB and no cores, and ran llama.cpp's prompt cache, which grew one to
+  3.2 GB. Without cores of its own, a judgment on a busy node missed its 5 s budget, and `fail: closed` refused the
+  request. It now gets what `server init` gives its judges: 2 GB, two cores each (fewer on a small node), and no
+  prompt cache. Every judge served through `modelserve` runs with `--cache-ram 0`.
+
+## v0.14.9 — 2026-10-19
+
+Sign-in set up from the console, with Active Directory tried before it is
+kept. A node's console now shows what the machine is doing, and a node's web
+page shows what it runs. Also fixes found trying Active Directory against a
+Samba domain controller on the test cluster. Sign-in through it, nested
+groups and a disabled account suspended by the sync all worked there.
+
+**Upgrading.**
+- No migration.
+- `auth.ldap` and `auth.oidc` are in the cluster's settings, as before. A
+  block written into `server.yaml` on a cluster with a store was never
+  read; the log said so. Put it in with `llmgw ctl config edit`, or set it up
+  again from the console's Sign-in screen.
+- A Samba DC's own autogenerated certificate is refused (a negative serial
+  and no SAN). Give it one from a CA.
+
+### Added
+
+- **Sign-in set up from the console.** Under People → Sign-in, `sudo llmgw` shows how people get in: single sign-on,
+  Active Directory and SCIM, each on or off. It also shows the directory's last sync. **E** sets up Entra ID from the
+  tenant and client IDs. **D** sets up Active Directory, and the directory is tried before it is kept: each server
+  reached and bound, how many people may sign in, how many are in each group. Nothing is kept if a step fails. The
+  password, client secret and SCIM token are kept in the store as `store:auth/…`, sealed by the master key. The
+  domain's CA goes into the settings (`auth.ldap.ca`), so no file has to be put on each control plane. The admin API
+  has the same: `GET /api/v1/auth`, `POST /api/v1/auth/ldap/test` and `PUT /api/v1/auth/secrets/{name}`.
+- **A node's page says what it runs.** It now shows the agent's version, marked when the control plane runs another,
+  and the engines with their versions. Each GPU row now names the models loaded on that card, and each engine row its
+  GPUs.
+- **`sudo llmgw` on a node opens on what the machine is doing.** A new Overview screen shows the agent's state and
+  version, each control plane it dials and whether it answers, every GPU with the models on it, the engines running
+  here, the free space for weights and engines, and the agent's warnings of the last hour. It is read on the node
+  itself, every three seconds, so it needs no admin token and works while the control plane is out of reach.
+
+### Fixed
+
+- **A broken `auth.ldap` no longer stops every control plane.** It is the cluster's setting. A CA file that one
+  machine could not use made `llmgw server` exit on every control plane, and no `ctl config` could then put it right.
+  Directory sign-in is now turned off on that control plane, with an error in its log. Local accounts still sign in.
+- **A second `llmgw server` on a machine stops before it competes to lead.** `llmgw server check` (no such command)
+  ran a whole server. Its ports were taken, yet it went on to hold the leader's lease for three hours while the real
+  control planes stood by. `server` now refuses words it does not know and takes its ports before anything else.
+- **The single sign-on and Active Directory guides said `server.yaml`.** On a cluster with a store, `auth` is in the
+  cluster's settings and the file's block is not read. They now say `llmgw ctl config edit`, and what certificate a
+  domain controller needs.
+
+## v0.14.8 — 2026-10-18
+
+Sign-in with an on-premises Active Directory, and Entra ID made safe for
+people in many groups. Also several keys for one outside provider, and a
+measure of what the gateway adds to a request.
+
+**Upgrading.**
+- Migration 0048 adds the `user_directory` table, applied when the control
+  plane starts.
+- Nothing changes until `auth.ldap` is set; docs/install.md, "Active
+  Directory", has an example. It has been tested against a fake directory,
+  not yet against a real domain controller.
+- A judge installed by an earlier `server init` is refitted by the leader
+  by itself (below); the steps given for v0.14.7 are no longer needed.
+
+### Added
+
+- **Sign-in with Active Directory** (`auth.ldap`). People sign in on the gateway's form with their AD username or mail
+  and password, checked against the domain controllers over LDAPS or StartTLS. Groups, nested ones included, give
+  staff roles and places in the tree, as with OIDC. Accounts are tied to the `objectGUID` and made at first sign-in
+  when `create_accounts` is on. Every 15 minutes the leader reads the directory. People disabled or deleted there are
+  suspended and reinstated when they come back; roles and memberships follow their groups. A sync that would suspend
+  more than half the accounts is refused. See docs/install.md, "Active Directory".
+
+- **Several keys for one outside provider.** An external model takes `api_key_files`, each with an optional weight, in
+  place of `api_key_file`. Each key is an endpoint of its own (`external:MODEL#1`, `#2`), so a key answering 429 is
+  passed over while its breaker is open and the request goes to another. A key answering 401 is passed over too, and
+  `external_key_refused` is sent, naming the endpoint, never the key. Usage records which key answered by that name.
+
+- **`llmgw ctl bench overhead`**: what the gateway adds to a request, measured against an engine that answers at
+  once. Gateway and engine run in one process on loopback, so no cluster is needed. The load is sent straight to the
+  engine and through the gateway: bare, with authentication and accounting, and with the guardrails' pattern checks.
+  It reports p50 and p99 latency, what the gateway added, and a rough CPU per thousand requests. On an Apple M5 the
+  gateway added 0.05–0.42 ms at p50 and at most 0.49 ms at p99 (README).
+
+### Fixed
+
+- **Entra ID's groups past 200 no longer take a role away.** For someone in more than 200 groups Entra sends none in
+  the token, only a pointer to Graph. That was read as no groups, so an administrator lost their role at the next
+  sign-in. Their role and memberships are now left as they were, and an error is logged saying what to change in the app
+  registration. docs/install.md has an Entra ID section.
+- **An account made by SCIM or OIDC could not be reinstated.** Setting it active was refused as "never approved".
+- **An upgrade refits the judge an earlier `server init` installed.** Upgrading to v0.14.7 left such a judge as it was:
+  no cores or one, 2 or 4 GB, the prompt cache on. On fewer than eight cores every request was refused until it was
+  fixed by hand. Now the leader finds a judge whose entry is exactly what an earlier init wrote and gives it 2 GB, no
+  prompt cache and cores by today's rule (up to four for one judge, two each for more). It then restarts the judge's
+  replicas onto that. The number of replicas is left as it is. A judge whose entry was changed by hand is left alone,
+  and the log says so once.
+- **A restart is no longer refused for room another restart is giving back.** With every core of a node reserved, two
+  models restarted together, one to need more and one less (as after `ctl model resources` on a judge and a starter),
+  had the first refused: "fits on no node even with this replica stopped", though the second was about to give back
+  its cores. A restart now waits for other restarts under way, up to the drain grace, before it is refused. A drained
+  replica also counts a replica being restarted that still answers as serving, so replicas restarted one after
+  another no longer wait out the five-minute grace each.
+
+## v0.14.7 — 2026-10-16
+
+The judge installed by `server init` sized for what it does. On a machine
+of fewer than eight cores v0.14.6 left every judgment past its budget, so
+`fail: closed` refused every request. A single-machine install of v0.14.7
+on a 4-core VM passed 20 of 20, and on 8 cores 20 of 20 under load.
+
+**Upgrading.**
+- No migration.
+- **Upgrading does not change a judge already in the catalog.** v0.14.6
+  and earlier installed two judges, reserving no cores or one, and 2 or
+  4 GB. To give them what they need:
+  `llmgw ctl model resources qwen3guard-0.6b --memory 2 --cores 2`, then
+  `llmgw ctl restart --model qwen3guard-0.6b`. The two then need four cores
+  on a single machine; below that, a longer `latency_budget` or
+  `fail: open`.
+- On a node whose cores are all reserved, `ctl restart --model` cannot
+  start the new replica beside the old one and waits. Stop the old engine
+  process, and the replica comes back with the new figures.
+
+### Fixed
+
+- **Both judges fit an 8 GB machine again.** v0.14.6 had `server init` reserve 4 GB a judge replica. That figure came
+  from one replica at 3.2 GB after a day of traffic. The growth was llama.cpp's prompt cache, up to 8 GB by default,
+  which a judge never reuses: each text it reads is different. On a 4-core, 8 GB machine the second judge then fitted
+  nowhere ("has 7 GB of memory with 1 free, needs 4"). The judge now runs with `--cache-ram 0`, uses about 1.6 GB and
+  stays there, and reserves 2.
+- **One machine runs one judge, on up to four cores.** A second replica on the same machine shares its cores and is
+  lost with it, so it bought nothing but half the speed of each. `server init` on a single machine now serves one judge
+  on up to four cores (2.3 s a judgment where two on two cores took 4.2). Where there will be other nodes, it serves two
+  on two cores each, as in HA setups. `--judge-replicas` and the setup's Copies still choose.
+- **A judge gets the two cores it needs.** v0.14.6 gave each judge replica one core on machines under eight. A judgment
+  of about 280 tokens takes 8.0 s on one core, 4.2 s on two and 2.3 s on four, against the 5 s budget, so on a 4-core
+  machine every judgment timed out and every request was refused. Where there are two judges, each now reserves two
+  cores (one only below four CPUs), and the one judge of a single machine up to four; the starter has what is left, or
+  shares when nothing is. A judge on two cores stays close to its budget: give it more cores or a GPU, or raise
+  `latency_budget`, where requests must not wait.
+
+## v0.14.6 — 2026-10-16
+
+Setup that says what each part of an install is and lets every one be
+chosen and set; three machines that each run the gateway, so any one can
+be lost; a cluster that loses its store's primary without losing a
+request; one ComfyUI for an environment's models; pictures, 3D and
+workflows as MCP tools; alerts asked for at setup; extended thinking.
+
+**Upgrading.**
+- No migration.
+- **Control planes**: with `control.lease_ttl` unset the lease is now 10 s
+  (was 15). Upgrade the standbys first and the leader last; a mix of
+  versions meanwhile is fine.
+- **The store** (PostgreSQL): each member is dialled for 3 s before the
+  next is tried. A store so far away that connecting takes longer needs
+  `connect_timeout=10` in its address.
+- **Nodes** notice a control plane that died without closing its
+  connection in 8 s (was 15) once they run this release.
+- **Judges installed by an earlier `server init`** keep reserving no cores
+  and 2 GB. To give them their own:
+  `llmgw ctl model resources qwen3guard-0.6b --memory 4 --cores 2`, then
+  `llmgw ctl restart --model qwen3guard-0.6b`.
+- **Patroni** replaces a lost primary within 20 s on a new cluster. On an
+  existing one, run `server init --patroni ...` again from the machine with
+  the authority; it changes only that.
+- **A floating address** keeps VRRP router id 51 until `server add --vip`
+  runs again, which moves it to one of its own; run it for every standby.
+- A request whose judge is not there yet waits up to the policy's
+  `latency_budget` for one before `fail: closed` refuses it.
+
+### Added
+
+- **Setup lets every part be chosen, each with its settings, and says what each is.** `sudo llmgw` on a new machine lists
+  the parts of an install: the control plane, the store, the model mirror, this machine as a node, the starter model,
+  pictures and 3D, the guardrails' judge, the firewall, notifications, limits and keys, on-demand turns, an offline
+  bundle. Beside each is what it will be or what is missing without it. Space turns a part on or off, and Enter opens
+  its settings in a dialog that explains it at more length. The settings are: which starter model (six Qwen sizes, or
+  any GGUF on Hugging Face) and its context; the mirror's directory or another machine's mirror; the judge's
+  categories, copies, wait and what happens when it does not answer; the portal's language; the sign-up form; the
+  default limits; the longest a key lives; the on-demand waits. A plan page then lists what is installed and what is
+  left out, with the same install as a command. Every setting is a `server init` flag too (`--starter`,
+  `--judge-fail`, `--media`, `--max-key-ttl`, ...).
+- **Three machines are high availability for the gateway, not only the store.** With three machines, setup makes the two
+  database machines control planes too, standing by while the first leads, with an optional floating address
+  (keepalived) that follows whichever leads. Any one machine can now be lost. By command: `server init --patroni ...
+  --standbys --vip 10.0.0.160/24`. The mirror's management API answers the three control planes, so a standby that
+  comes to lead can still fetch models.
+
 - **Pictures uploaded to work from**: `/v1/files` takes `purpose=vision` or
   `user_data` besides `batch`, kept as long as job results, so a picture is
   sent once and named by its `file-…` id in the MCP tools and a workflow's
@@ -71,6 +1004,42 @@ APIs change between minors without a deprecation period.
 
 ### Fixed
 
+- **The judges and the starter model divide a machine's CPUs instead of each taking all of them.** `server init` gave the
+  judge and the starter no cores of their own, so every llama.cpp engine ran on every core: two judges and the starter
+  on a 4-core control plane competed, and under load 38 of 60 requests were refused for want of a judgment. Moved to a
+  node of their own with 2 cores each, the same test refused 2. When this machine is the node, init now gives each judge
+  replica 1 core (2 from 8 cores) and the starter what is left (at most 8); judges bound for other nodes get 2 each. A
+  judge also reserves 4 GB, not 2: it uses about 3.2.
+
+- **A lost store machine stops costing requests most of a minute.** The store's connections try each member for 3 seconds,
+  not until the kernel gives up, and a connection to a machine that is gone is closed within about 10 seconds (TCP
+  keepalive, and `TCP_USER_TIMEOUT` on Linux). Patroni replaces a lost primary within 20 seconds (`ttl` 20, `loop_wait`
+  5, `retry_timeout` 5). Running `server init --patroni` again applies this to an existing cluster. Measured on three
+  VMs with the primary's power cut: about 45 seconds of hanging requests before.
+- **Leadership stays put while the store's primary is replaced.** Replacing a lost primary takes longer than the
+  leadership lease lives, so the lease had lapsed for everyone by the time the store answered again. Whoever asked first
+  took it, as often a standby as the leader. A standby that saw the store out of reach now leaves the lease alone for one
+  round after it returns, so the leader renews it, and a leader that is truly gone is replaced a round later. With that,
+  the default lease is 10 seconds (was 15). A node notices a control plane that died without closing its connection in
+  8 seconds (was 15), so it is on its way to the new leader as that one takes over.
+- **A judge replica that hangs no longer costs the request.** All the tries of a judge request shared the policy's
+  budget, so a replica that hung used it all and the try on another replica had none left: under `fail: closed` the
+  request was refused while a judge stood idle. When the first replica has not answered by half the budget, the same
+  question now goes to a replica on another machine too, and the first answer is taken. The first replica is not cut
+  short, which on a slow machine would refuse answers it was about to give. Replicas on one machine whose CPUs are taken
+  are slow together, so asking both only doubles the work there: measured that way, as many requests failed as before.
+  On such a machine the remedy is room, not a second question: judges on more than one node, more cores, or the judge
+  on a GPU.
+- **A new leader waits for the judge instead of refusing at once.** The moment a control plane takes the lead it knows no
+  judge yet. When every judge fails at once, they are asked again every 250 ms while the policy's `latency_budget` lasts,
+  so `fail: closed` no longer answers every request in those seconds with 503 (about 17 seconds of them in the test). A
+  judge that is slow is not asked again, and an alarm is raised once per request, not once per try.
+
+- Node discovery for Prometheus listed only the control plane asked and the
+  store, so a machine that is both was a control plane to one and a store
+  member to the other — scraped twice — and a standby asked alone lost every
+  node. Every control plane lists all of them alike; deploy/monitoring's
+  prometheus.yml says to scrape and ask each.
 - A model whose weights and context pool did not fit its card was added
   and then failed to make its context on every start, taking the card from
   others each time (Apertus 70B with a 65536-token pool, 8B with 131072, on

LICENSE

@@ -0,0 +1,105 @@
+# Functional Source License, Version 1.1, ALv2 Future License
+
+## Abbreviation
+
+FSL-1.1-ALv2
+
+## Notice
+
+Copyright 2026 Ilker Manap
+
+## Terms and Conditions
+
+### Licensor ("We")
+
+The party offering the Software under these Terms and Conditions.
+
+### The Software
+
+The "Software" is each version of the software that we make available under
+these Terms and Conditions, as indicated by our inclusion of these Terms and
+Conditions with the Software.
+
+### License Grant
+
+Subject to your compliance with this License Grant and the Patents,
+Redistribution and Trademark clauses below, we hereby grant you the right to
+use, copy, modify, create derivative works, publicly perform, publicly display
+and redistribute the Software for any Permitted Purpose identified below.
+
+### Permitted Purpose
+
+A Permitted Purpose is any purpose other than a Competing Use. A Competing Use
+means making the Software available to others in a commercial product or
+service that:
+
+1. substitutes for the Software;
+
+2. substitutes for any other product or service we offer using the Software
+   that exists as of the date we make the Software available; or
+
+3. offers the same or substantially similar functionality as the Software.
+
+Permitted Purposes specifically include using the Software:
+
+1. for your internal use and access;
+
+2. for non-commercial education;
+
+3. for non-commercial research; and
+
+4. in connection with professional services that you provide to a licensee
+   using the Software in accordance with these Terms and Conditions.
+
+### Patents
+
+To the extent your use for a Permitted Purpose would necessarily infringe our
+patents, the license grant above includes a license under our patents. If you
+make a claim against any party that the Software infringes or contributes to
+the infringement of any patent, then your patent license to the Software ends
+immediately.
+
+### Redistribution
+
+The Terms and Conditions apply to all copies, modifications and derivatives of
+the Software.
+
+If you redistribute any copies, modifications or derivatives of the Software,
+you must include a copy of or a link to these Terms and Conditions and not
+remove any copyright notices provided in or with the Software.
+
+### Disclaimer
+
+THE SOFTWARE IS PROVIDED "AS IS" AND WITHOUT WARRANTIES OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING WITHOUT LIMITATION WARRANTIES OF FITNESS FOR A PARTICULAR
+PURPOSE, MERCHANTABILITY, TITLE OR NON-INFRINGEMENT.
+
+IN NO EVENT WILL WE HAVE ANY LIABILITY TO YOU ARISING OUT OF OR RELATED TO THE
+SOFTWARE, INCLUDING INDIRECT, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES,
+EVEN IF WE HAVE BEEN INFORMED OF THEIR POSSIBILITY IN ADVANCE.
+
+### Trademarks
+
+Except for displaying the License Details and identifying us as the origin of
+the Software, you have no right under these Terms and Conditions to use our
+trademarks, trade names, service marks or product names.
+
+## Grant of Future License
+
+We hereby irrevocably grant you an additional license to use the Software under
+the Apache License, Version 2.0 that is effective on the second anniversary of
+the date we make the Software available. On or after that date, you may use the
+Software under the Apache License, Version 2.0, in which case the following
+will apply:
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use
+this file except in compliance with the License.
+
+You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed
+under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
+CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.

README.md

@@ -40,7 +40,9 @@ Releases track the roadmap phases; see [CHANGELOG.md](CHANGELOG.md).
 | v0.11.0 | — | High availability from setup, the setup and console screens, usage sums and per-tenant alerts | released |
 | v0.12.0 | — | The cluster's settings in the store, in force when changed; the firewall switch; wipe | released |
 | v0.13.0 | — | The whole system stopped and started from one command; Qwen3Guard judges, on a GPU and CPUs together | released |
-| v0.14.0 | — | Knowledge bases: documents from people's own machines, kept apart per tenant, answered from | released (v0.14.1: adapters, the tree over the API, the firewall on every machine; v0.14.2: pictures from ComfyUI; v0.14.3: 3D models as jobs; v0.14.4: any ComfyUI workflow as a job, presets, licences; v0.14.5: picture and 3D models that add themselves) |
+| v0.14.0 | — | Knowledge bases: documents from people's own machines, kept apart per tenant, answered from | released (v0.14.1: adapters, the tree over the API, the firewall on every machine; v0.14.2: pictures from ComfyUI; v0.14.3: 3D models as jobs; v0.14.4: any ComfyUI workflow as a job, presets, licences; v0.14.5: picture and 3D models that add themselves; v0.14.6: setup that explains every part and lets it be chosen, the gateway on all three HA machines; v0.14.7: the judge sized for its machine; v0.14.8: sign-in with Active Directory, several keys for one outside provider; v0.14.9: sign-in set up from the console, a node's console that shows the machine) |
+| v0.15.0 | — | The company's MCP tools and VS Code through the gateway; guardrails that ask outside services and put redacted values back; answers to similar questions | released (v0.15.1: Strata, a 125B model of experts on one card; a second mirror kept in step; pickled weights refused; a served model's context changed; v0.15.2: a served model moved to more cards; v0.15.3: a model stopped and kept, or removed with its files) |
+| v0.16.0 | — | Weights pinned to a commit and checked against what Hugging Face published; documents embedded by the gateway; Strata offered where models are chosen; cluster upgrades that say how far they are | released (v0.16.1: the mirror's Strata tab, Strata's ready-made packs; v0.16.2: standbys that answer clients, FreeToken, a model fetched once per node; v0.16.3: replicas per profile, a served model's settings on the mirror page, the CPU's speed before serving) |
 | v1.0.0 | 5 | What phase 5 has left: precise KV routing proven on vLLM, the block cost picker, defragmentation across NVLink domains, prefill/decode disaggregation | planned |
 
 As of v0.6.0 a cluster runs unattended: agents report their hardware, the
@@ -50,6 +52,17 @@ budgets and rate limits decide who may use any of it.
 
 Since v0.11.0 (each release's detail is in [CHANGELOG.md](CHANGELOG.md)):
 
+- **Setup that explains, and high availability that is whole** (v0.14.6).
+  `sudo llmgw` lists the parts of an install, each with what it is for and
+  what is missing without it, and every one can be turned off or set: the
+  starter model and its context, the mirror's directory, the judge's
+  categories and what it does when it does not answer, limits, the
+  portal's language, pictures and 3D. With three machines each runs the
+  control plane, with a floating address on the leader, and any one can
+  be lost: on three VMs, cutting the store primary's power lost no
+  request. One ComfyUI serves an environment's models, pictures, 3D and
+  workflows are MCP tools, and `/v1/messages` gives a model's reasoning as
+  thinking blocks.
 - **Picture and 3D models that add themselves** (v0.14.5). Where a node's
   card fits flux-schnell or triposr and the catalog makes nothing of that
   kind, the leader adds it on demand, so a new cluster's portal has its
@@ -172,6 +185,8 @@ Every page is also in the binary: `llmgw docs`, or **Manual** in the console.
   cluster that exists
 - [docs/kubernetes.md](docs/kubernetes.md) — the control plane on
   Kubernetes with the Helm chart, and why GPU nodes stay outside it
+- [docs/kong.md](docs/kong.md) — llmgw behind Kong Gateway, or Kong behind
+  llmgw: the configuration, what Kong counts, and which to choose
 - [docs/fabrics.md](docs/fabrics.md) — InfiniBand fabrics and NVLink: what
   llmgw needs of them, checking a node with `llmgw doctor`, and what to look at
   when a replica spanning nodes will not run
@@ -246,6 +261,33 @@ Add `--json` to any command for machine-readable output.
 says nothing about which of six constraints bit, so it prints every candidate
 considered and the reason it was rejected.
 
+## What llmgw adds to a request
+
+Measured with `llmgw ctl bench overhead`. It starts an engine that answers
+at once and a gateway in front of it, in one process on loopback. It sends
+the same load straight to the engine and through the gateway, and reports
+what the gateway added. Run on an Apple M5 (10 CPUs, load about 2) on
+2026-10-16, 10 s a step, 72,000 requests with none failed:
+
+| Through the gateway | added at p50 | added at p99 |
+|---|---|---|
+| bare | 0.06–0.24 ms | 0.10–0.33 ms |
+| with authentication and usage accounting | 0.05–0.32 ms | 0.09–0.37 ms |
+| with the guardrails' pattern checks as well | 0.10–0.42 ms | 0.15–0.49 ms |
+
+The range is across 100 to 2,000 requests a second. The overhead was
+*lowest* at 2,000, most likely because a busier processor keeps its cores
+awake. The guardrails' judge model is not in these figures: a judgment
+takes seconds on CPUs (see [install.md](docs/install.md)) and is a model's
+time, not the gateway's.
+
+The command also prints the gateway's CPU per thousand requests. Gateway,
+engine and load share one process there, so that figure is rough (0.04 to
+0.65 s per thousand in the run above); the latency is the one to quote.
+Figures from a VM on a busy host move with the host: run it where you will
+deploy, and keep the load it prints beside the result. CI runs a one-second
+version on every change (`TestTheGatewaysOverheadIsMeasured`).
+
 ## Running the router
 
 ```sh
@@ -273,8 +315,9 @@ within 1.24x of idle while 500 concurrent batch requests run, against 58x with
 admission off.
 
 Routing profiles select how endpoints are scored: `chat` favours session
-affinity, `batch` favours the shortest queue, `long-context` favours free KV
-cache. Pick one per model in the config, or per request with the
+affinity and the replica that has been answering soonest (its time to first
+token, smoothed), `batch` favours the shortest queue, `long-context` favours
+free KV cache. Pick one per model in the config, or per request with the
 `x-llmgw-profile` header.
 
 ## sim-engine
@@ -329,3 +372,10 @@ make rename NEW=git.example.com/team/llmgw
 ## Language
 
 Code, comments, tests, commit messages and documentation are in English.
+
+## Licence
+
+Functional Source License 1.1 with an Apache 2.0 future licence
+(FSL-1.1-ALv2); see [LICENSE](LICENSE). Each release may be used, read and
+changed for anything except offering it as a competing product or service,
+and becomes Apache 2.0 two years after it is published.

api/proto/llmgw/v1/agent.proto

@@ -30,6 +30,7 @@ message AgentMessage {
     KVEvents kv_events = 8;
     TrainStatus train_status = 9;
     TrainArtifact train_artifact = 10;
+    SlurmStatus slurm_status = 11;
   }
 }
 
@@ -78,6 +79,7 @@ message ControlMessage {
     AgentSettings agent_settings = 15;
     TrainStart train_start = 16;
     TrainCancel train_cancel = 17;
+    Slurm slurm = 18;
   }
   // trace is the W3C trace context (traceparent, and tracestate when there
   // is one) of the control plane's span that sent this command, so what the
@@ -238,6 +240,13 @@ message Host {
   // nvidia_driver is the driver version, empty when there is none. A card
   // without a driver is not a GPU anything can use.
   string nvidia_driver = 6;
+  // mem_bandwidth_gbs is how fast the memory reads, in GB/s, measured when
+  // the agent started: what decides a model's speed on the CPU. 0 unknown.
+  double mem_bandwidth_gbs = 7;
+  // mem_bandwidth_core_gbs is what one core reads alone, in GB/s: a replica
+  // given a few threads of a large machine reads at a few times this, not at
+  // the whole machine's rate. 0 unknown.
+  double mem_bandwidth_core_gbs = 8;
 }
 
 // Storage is one filesystem a node keeps weights on.
@@ -334,6 +343,13 @@ message GPU {
   // memory the instance's share; utilization_percent stays zero, because no
   // driver measures it per instance.
   MIGDevice mig = 22;
+  // How the card is linked to its host: the fastest generation card and
+  // slot allow together, the card's own fastest, and the lanes in use and
+  // the card's most. Zero when the driver did not say.
+  int32 pcie_gen = 23;
+  int32 pcie_gpu_gen = 24;
+  int32 pcie_width = 25;
+  int32 pcie_width_max = 26;
 }
 
 // MIGDevice is which card a MIG instance is carved from, and how.
@@ -438,6 +454,21 @@ message GPUSensor {
   int32 last_xid = 26;
   string last_xid_text = 27;
   int64 last_xid_unix = 28;
+  // How much of the chip the work on it uses, which utilization_percent
+  // does not say: that is the share of time any kernel ran, and a kernel on
+  // one SM of a hundred counts in full. has_profile says these are present;
+  // -1 is not reported. Read from NVML's GPM on Hopper and later, or from a
+  // dcgm-exporter on the node; profile_source says which ("gpm", "dcgm").
+  bool has_profile = 29;
+  // sm_active_percent is the share of time the SMs had work, averaged over
+  // all of them; sm_occupancy_percent how full those SMs were with warps.
+  double sm_active_percent = 30;
+  double sm_occupancy_percent = 31;
+  // tensor_active_percent is the tensor cores; dram_active_percent the
+  // memory interface, which decode is bound by.
+  double tensor_active_percent = 32;
+  double dram_active_percent = 33;
+  string profile_source = 34;
 }
 
 // ShardLoad is one engine process group's use.
@@ -588,6 +619,13 @@ message Heartbeat {
   // is not handed out. Empty from an agent that does not measure it, or from
   // one that could not this time.
   repeated int32 gpu_foreign_memory_mb = 19;
+
+  // mem_bandwidth_gbs is how fast the node's memory reads, in GB/s, once
+  // the agent has measured it — in the background after it starts, so a
+  // registration usually has none yet. 0 means not measured.
+  double mem_bandwidth_gbs = 20;
+  // mem_bandwidth_core_gbs is one core's read rate, as in Host.
+  double mem_bandwidth_core_gbs = 21;
 }
 
 // Provision is the software a node should have: which engines, at which
@@ -690,6 +728,9 @@ message FetchProgress {
   // is a real answer: a percentage would then be invented, and an invented
   // percentage that stops moving is worse than a byte count that keeps rising.
   int64 bytes_total = 3;
+  // phase is what the bytes are: empty while they arrive, "checking" while the
+  // node checks a set's files against the mirror's checksums.
+  string phase = 4;
 }
 
 // StartShard launches one shard. It is idempotent by shard_id: the control
@@ -749,6 +790,13 @@ message EngineProfile {
   // kv_events has a vLLM engine publish its KV cache events to the agent, for
   // the router's precise index.
   bool kv_events = 13;
+  // allow_pickle lets the node fetch weights in a pickle format, which run
+  // code when loaded: the catalog's allow_pickle, set on purpose.
+  bool allow_pickle = 14;
+  // model_sha256 is the sha256 the registry published for the file
+  // model_path names, from the catalog's origin: the node refuses a copy
+  // that is not it. Empty when nothing is known to check against.
+  string model_sha256 = 15;
 }
 
 // RemoveWeights deletes cached model files a node no longer needs.
@@ -940,3 +988,99 @@ message TrainArtifact {
   bytes data = 4;
   bool last = 5;
 }
+
+// Slurm is what this node is in the Slurm cluster run beside llmgw
+// (BACKLOG P): its controller, one of its nodes, both or neither. Sent when
+// an agent registers and whenever it changes; the agent installs what is
+// missing and keeps the daemons running as it says.
+message Slurm {
+  // controller runs slurmctld here, on files.
+  bool controller = 1;
+  // node runs slurmd here, configless, fetching its files from conf_server.
+  bool node = 2;
+  // name is this node's NodeName in slurm.conf.
+  string name = 3;
+  // conf_server is slurmctld's host:port.
+  string conf_server = 4;
+  // munge_key is the key every machine of the cluster shares.
+  bytes munge_key = 5;
+  // files are the controller's, by name in /etc/slurm; empty for a node
+  // that is not the controller.
+  map<string, string> files = 6;
+  // lent says Slurm may run jobs here now: resume the node if llmgw drained
+  // it. Otherwise it is drained, its jobs left to finish.
+  bool lent = 7;
+  // shared is the file system every machine of the Slurm cluster sees at
+  // the same path; unset or with no path, none.
+  SlurmShared shared = 8;
+  // manage_users has llmgw keep the Linux users below on this machine,
+  // and take away those it made that are no longer among them.
+  bool manage_users = 9;
+  repeated SlurmUser users = 10;
+  // make_homes has this machine make the users' home directories: the one
+  // that exports the shared path, else the controller.
+  bool make_homes = 11;
+  // containers has the machine install pyxis and enroot, so jobs may run
+  // in containers (srun --container-image), and set them up.
+  bool containers = 12;
+}
+
+// SlurmShared is where the Slurm cluster's shared files come from.
+message SlurmShared {
+  // path is where every machine sees them, as /shared.
+  string path = 1;
+  // nfs is an export, host:/path, to mount at path; empty when this
+  // machine exports path itself (serve) or the site mounts it.
+  string nfs = 2;
+  // serve has this machine export path over NFS to clients.
+  bool serve = 3;
+  // clients are the addresses the export answers.
+  repeated string clients = 4;
+}
+
+// SlurmUser is a Linux user that runs jobs, the same on every machine.
+message SlurmUser {
+  string login = 1;
+  // uid is also the user's own group's gid.
+  uint32 uid = 2;
+  // name is the account's, for the GECOS field.
+  string name = 3;
+  string home = 4;
+  // ssh_keys are OpenSSH public keys, one a line of authorized_keys.
+  repeated string ssh_keys = 5;
+}
+
+// SlurmStatus is what this node's Slurm daemons are doing, sent when it
+// changes and every few minutes while the node has a part in Slurm.
+message SlurmStatus {
+  // version is Slurm's, as "23.11.4"; empty when it is not installed.
+  string version = 1;
+  // hardware is `slurmd -C` on a node, which slurm.conf must match.
+  string hardware = 2;
+  bool slurmd_running = 3;
+  bool slurmctld_running = 4;
+  // state and reason are the controller's for this node, from scontrol.
+  string state = 5;
+  string reason = 6;
+  // jobs is how many job steps run on this node now, counted from their
+  // slurmstepd so a controller that is down does not hide them; -1 when the
+  // agent cannot tell.
+  int32 jobs = 7;
+  // error is the last thing that went wrong keeping Slurm, empty when
+  // nothing did.
+  string error = 8;
+  // files_sha256 is of the controller's files as written, so the leader
+  // knows they arrived.
+  string files_sha256 = 9;
+  // slurmdbd_running is the accounting daemon on the controller, which keeps
+  // the jobs' records in a MariaDB beside it.
+  bool slurmdbd_running = 10;
+  // shared is the shared path as this machine has it: "mounted",
+  // "exported", "missing", or empty when there is none.
+  string shared = 11;
+  // users is how many of llmgw's users this machine has.
+  int32 users = 12;
+  // containers is what this machine has for jobs in containers, as
+  // "pyxis 0.24.0, enroot 4.2.1"; empty while it has not got both.
+  string containers = 13;
+}

cmd/llmgw-kb/main.go

@@ -331,6 +331,8 @@ hidden files, version control folders and Office lock files are not.`,
 	f.BoolVar(&o.Delete, "delete", false, "remove documents this machine sent whose files are gone")
 	f.BoolVar(&o.Retry, "retry", false, "read again files skipped before")
 	f.IntVar(&o.Workers, "workers", 4, "files read and sent at once")
+	f.BoolVar(&o.EmbedHere, "embed-here", false,
+		"embed on this machine and send the vectors, even to a gateway that embeds text it is sent")
 	f.StringVar(&ocr, "ocr-languages", "tur+eng", "Tesseract's languages for images and scanned pages")
 	f.Int64Var(&maxMiB, "max-size", 256, "skip files larger than this many MiB")
 	f.BoolVarP(&verbose, "verbose", "v", false, "list unchanged files too")

deploy/ha/README.md

@@ -1,4 +1,4 @@
-# Two control planes
+# More than one control plane
 
 llmgw handles the hard half of this itself: only one control plane places work,
 and agents find whichever one is leading. What it does not do is move client
@@ -37,7 +37,7 @@ nothing in front of the two had moved.
 So something has to. The options, in the order most people should consider
 them:
 
-**A load balancer you already run.** Health-check `/ready` on each control
+**A load balancer you already run.** Health-check `/leader` on each control
 plane and send traffic to whichever answers. This is the right answer if you
 have one, and it needs nothing else from this directory.
 
@@ -50,25 +50,50 @@ in whatever your resolvers actually honour rather than what you set, which on
 a bad day is minutes. Fine for a cluster people use interactively, wrong for
 one an agent is driving.
 
-## Check `/ready`, not `/health`
+## Check `/leader`, not `/health`
 
 This is the one thing to get right, and the obvious choice is the wrong one.
 
-`/health` says the process is up. It says nothing about whether it can serve,
+`/health` says the process is up. It says nothing about whether it leads,
 and a standby is exactly that case: alive, answering, and holding no endpoints
 at all — its reconciler placed nothing and it has no agent streams to adopt
-from. A load balancer checking `/health` would send it half the traffic, and
-half the traffic would be a 503.
+from. A standby passes client requests on to the leader, so traffic sent to
+one still works — until the leader is the machine that was lost, and there
+is nobody to pass them to.
+
+`/leader` is 200 on the control plane that leads and 503 on a standby. It is
+what keepalived follows for the floating address, and what a load balancer
+should send traffic by: to the one machine that answers 200.
+
+`/ready` is narrower: 200 only when this process has somewhere to send a
+request, so 503 on a standby and also on a leader with no model running yet.
+Checking it takes the portal and the admin pages away with the models — on a
+new cluster with no node yet, nobody can sign in to add one. Use it to stop
+sending inference only, if your load balancer can tell the paths apart.
 
-`/ready` says this process has somewhere to send a request. It is 503 on a
-standby, 503 on a leader whose replicas have all failed, and 200 when there is
-something to serve — which is the question a load balancer is actually asking.
 Both are open, because a load balancer has no key.
 
-During a failover neither passes for about the length of the lease handover:
+During a failover nothing passes for about the length of the lease handover:
 three seconds when the leader is stopped cleanly, up to the lease TTL when it
-is not. That gap is real and there is no way to have it both ways — traffic
-cannot go to a control plane that has nothing to route to.
+is not. A proxy that waits that long for a healthy machine rather than
+failing at once (Caddy's `lb_try_duration`) hides it from clients.
+
+**Caddy**, in front of three control planes — the end of
+[deploy/proxy/Caddyfile](../proxy/Caddyfile) has it with the rest of the
+settings:
+
+```
+reverse_proxy 10.0.0.157:8080 10.0.0.158:8080 10.0.0.159:8080 {
+	lb_policy first
+	health_uri /leader
+	health_interval 2s
+	lb_try_duration 15s
+	flush_interval -1
+}
+```
+
+With a floating address the address already follows the leader: give Caddy
+that one address instead.
 
 Not the admin port, for a different reason: it serves pprof, and pprof can
 dump the process heap, where API keys sit in plaintext.

deploy/ha/keepalived.conf.example

@@ -41,7 +41,9 @@ vrrp_instance llmgw {
     # MASTER on the machine you would rather serve; BACKUP on the other.
     state MASTER
     interface eth0
-    # Any number both machines agree on and nothing else on the segment uses.
+    # Any number both machines agree on and nothing else on the segment uses:
+    # another cluster's keepalived with the same one reads as an invalid
+    # password. `llmgw server add --vip` derives one from the address.
     virtual_router_id 51
     # The higher one wins while both are healthy. The weight above is enough
     # to invert this when the preferred machine's gateway is not answering.

deploy/monitoring/grafana/llmgw-overview.json

@@ -504,7 +504,7 @@
     {
       "type": "timeseries",
       "title": "GPU busy, by card",
-      "description": "Share of each sample period a kernel was running, sampled every heartbeat.",
+      "description": "Share of each sample period in which any kernel ran (nvidia-smi's utilization.gpu), sampled every heartbeat. Not how much of the chip: one SM busy reads 100%. The SMs panel and power against its limit say how much.",
       "datasource": {
         "type": "prometheus",
         "uid": "llmgw-prom"
@@ -908,6 +908,51 @@
           "mode": "multi"
         }
       }
+    },
+    {
+      "type": "timeseries",
+      "title": "GPU SMs and tensor cores active, by card",
+      "description": "From the cards' profiling counters: NVML's GPM on Hopper and later, or a dcgm-exporter on the node. SMs active is the share of time the SMs had work, averaged over all of them; the gap to busy is chip a running kernel leaves idle. Tensor cores stay low in decode at batch one; prefill and batches are where they rise. Empty for cards nothing can profile.",
+      "datasource": {
+        "type": "prometheus",
+        "uid": "llmgw-prom"
+      },
+      "gridPos": {
+        "x": 0,
+        "y": 73,
+        "w": 24,
+        "h": 7
+      },
+      "id": 25,
+      "targets": [
+        {
+          "expr": "llmgw_gpu_sm_active_percent",
+          "legendFormat": "{{node}} gpu {{gpu}} SMs",
+          "refId": "A"
+        },
+        {
+          "expr": "llmgw_gpu_tensor_active_percent",
+          "legendFormat": "{{node}} gpu {{gpu}} tensor",
+          "refId": "B"
+        }
+      ],
+      "fieldConfig": {
+        "defaults": {
+          "unit": "percent",
+          "min": 0,
+          "max": 100
+        },
+        "overrides": []
+      },
+      "options": {
+        "legend": {
+          "displayMode": "list",
+          "placement": "bottom"
+        },
+        "tooltip": {
+          "mode": "multi"
+        }
+      }
     }
   ]
 }

deploy/monitoring/prometheus.yml

@@ -21,11 +21,18 @@ scrape_configs:
   # The control plane's dedicated metrics port. Not its admin listener: that one
   # is loopback because it also serves pprof, and pprof can dump the heap where
   # API keys sit in plaintext.
+  #
+  # Every control plane, one line each: the leader moves on restarts and the
+  # router's metrics are the leading one's, so scraping one alone leaves gaps
+  # — and alerts on them — whenever another leads.
   - job_name: llmgw
     static_configs:
       - targets: ['10.0.0.140:9091']
         labels:
-          instance: control-plane
+          instance: control-plane-140
+      # - targets: ['10.0.0.150:9091']
+      #   labels:
+      #     instance: control-plane-150
 
   # Every machine of the cluster, for its node exporter: network and disk
   # traffic, filesystems, CPU time. The control plane lists them — nodes,
@@ -33,10 +40,15 @@ scrape_configs:
   # nothing edited here, and each carries its llmgw id as the node label.
   # `llmgw node add` installs the exporter; on a machine added before that,
   # apt-get install prometheus-node-exporter.
+  # Asked of every control plane too: a standby lists only the control
+  # planes and the store, and they label every machine alike, so the same
+  # target from two of them is scraped once.
   - job_name: nodes
     http_sd_configs:
       - url: http://10.0.0.140:9091/prometheus/targets
         refresh_interval: 1m
+      # - url: http://10.0.0.150:9091/prometheus/targets
+      #   refresh_interval: 1m
 
   - job_name: prometheus
     static_configs:

deploy/proxy/Caddyfile

@@ -14,6 +14,7 @@ llm.example.com {
 		max_size 32MB
 	}
 
+	# One control plane. With more than one, see the end of this file.
 	reverse_proxy 127.0.0.1:8080 {
 		# Tokens are written the moment they are generated. Without this a
 		# streamed reply can arrive in bursts, and a client measuring time to
@@ -39,3 +40,36 @@ llm.example.com {
 	# Access log to stderr, which under systemd is the journal.
 	log
 }
+
+# More than one control plane (server init --standbys, server add): one of
+# them leads, the others stand by. Name them all and send to the one that
+# leads — /leader is 200 there and 503 on a standby. Not /ready: that is 503
+# on a leader with no model running yet, and would take the portal away with
+# it. A standby would pass a request on to the leader anyway, but not while
+# there is none, and the hop is for nothing.
+#
+# With a floating address (--vip) the address already follows the leader:
+# give reverse_proxy that one address instead, and keep the rest as above.
+#
+# llm.example.com {
+# 	request_body {
+# 		max_size 32MB
+# 	}
+# 	reverse_proxy 10.0.0.157:8080 10.0.0.158:8080 10.0.0.159:8080 {
+# 		# The first that is healthy: the leader, whichever it is.
+# 		lb_policy first
+# 		health_uri /leader
+# 		health_interval 2s
+# 		health_timeout 2s
+# 		# While leadership moves — a few seconds when a leader is lost —
+# 		# none is healthy; a request waits that long for one rather than
+# 		# failing at once.
+# 		lb_try_duration 15s
+# 		lb_try_interval 250ms
+# 		flush_interval -1
+# 		transport http {
+# 			response_header_timeout 15m
+# 		}
+# 	}
+# 	log
+# }

deploy/systemd/llmgw-agent.service

@@ -2,6 +2,9 @@
 Description=llmgw node agent
 After=network-online.target
 Wants=network-online.target
+# Never given up on: a new release that fails its first starts, or an upgrade
+# that restarts it twice in a minute, would otherwise leave the node off.
+StartLimitIntervalSec=0
 
 [Service]
 ExecStart=/usr/local/bin/llmgw agent --config /etc/llmgw/agent.yaml

docs/client-api.openapi.json

@@ -2890,6 +2890,10 @@
           "writable": {
             "type": "boolean",
             "description": "Writable is whether this caller may add and remove documents."
+          },
+          "server_embeds": {
+            "type": "boolean",
+            "description": "ServerEmbeds says a document may be sent as text alone: the gateway embeds its chunks, as the caller, with the collection's model."
           }
         }
       },

docs/clients.md

@@ -579,6 +579,15 @@ claude mcp add --transport http llmgw https://llm.example.com/mcp \
   --header "Authorization: Bearer $LLMGW_API_KEY"
 ```
 
+**Your company's tools.** Where an administrator has registered the
+company's own MCP servers (Jira, a file server, a database) and your key is
+granted their tools, the same `/mcp` lists them too, each as `server.tool`
+(`jira.search`). Give your agent only llmgw's `/mcp` and your key: the
+servers' own credentials stay with llmgw. A tool your key is not granted is
+not listed, and calling it by name is refused. Each call is in the audit
+log under your key: which tool, the size of what was sent, how it ended and
+how long it took, never what it said.
+
 ### Licences you leave out
 
 A model's licence is in `/v1/models` (`license`: its `id`, `name`, whether

docs/console.md

@@ -2,8 +2,8 @@
 
 `sudo llmgw`, with nothing after it, on a machine llmgw runs on. On a machine
 with nothing installed it opens the setup screens; on a control plane, the
-console described here; on a node, the node's own console, which edits
-`agent.yaml`. Everything on these screens can also be done by command — each
+console described here; on a node, the node's own console, which shows
+what runs on that machine and edits `agent.yaml`. Everything on these screens can also be done by command — each
 screen says which — so a script never needs them and a person never needs
 to remember the commands.
 
@@ -67,7 +67,8 @@ or with the mirror unticked, is said in the form and nothing starts.
 ## Watch
 
 **Overview** is `llmgw ctl status`: nodes, GPUs by class, fabrics, replicas,
-models, the store's members, and anything parked, unplaced or idle.
+models, the store's members, and anything parked, unplaced or idle. With no
+webhook in the cluster's settings it says the alerts go nowhere.
 
 **Monitor** draws the cluster the way btop draws a machine: CPU, memory,
 GPU busy and GPU memory, requests and tokens a second, time to first token,
@@ -173,6 +174,23 @@ search…** as an administrator, or **Delete…** it. People
 fill collections from their own machines with `llmgw-kb`
 ([knowledge.md](knowledge.md)). Off on a SQLite store, and says why.
 
+**MCP servers** lists the company's own MCP servers that agents reach through
+the gateway's `/mcp`, their tools named `server.tool`: whether each is on,
+who has been granted its tools, its rate per caller, the names of the
+credentials it is sent (never their values) and its address. **A** adds one:
+its name, which is its tools' prefix, its address, and a bearer token or
+another header, kept sealed in the store. On a row: **Test, and list its
+tools**, with its credential, as a call would; **Grant its tools to…** a
+team, project, organization, person or key — all of them, `jira.*`, or some
+by name — beside whatever else that level is granted, and saying when a
+level above keeps them out of reach; **Revoke** a grant;
+**Change…** its address, timeout or rate; **Replace its credential…**;
+**Turn off** (its tools are listed nowhere and every call refused) or on;
+**Remove…**. A server's tools are seen nowhere until granted. The screen
+also says how a person points an agent at the gateway, with a key of their
+own from `/profile/keys`. `llmgw ctl mcp` does the same from a shell, and
+the web page **MCP servers** (`/admin/mcp`, administrators only) in a browser.
+
 **Mirror** is what the model mirror holds, one row for each version of each
 model — a file, or the set of files a model of several is: shards, a vision
 projector, a config and tokenizer — with how far its download has got and
@@ -190,6 +208,19 @@ starting — and whether it ended done, failed, or waiting for room on busy
 cards. The entry shows how many are running. Enter on one shows its steps.
 `llmgw ctl jobs` is the same, and `llmgw ctl jobs ID` follows one.
 
+**Slurm** is the Slurm cluster run beside llmgw, when there is one: its
+controller and whether slurmctld runs, and each node that runs slurmd —
+whose it is now (llmgw's, Slurm's, or on its way between them), its state in
+Slurm, the jobs running there. On a row, **Lend to Slurm…** moves the node's
+replicas off and then lets Slurm run jobs there; **Take back from Slurm…**
+starts no new job there and gives it back to llmgw when its jobs have ended.
+Both change `slurm.lent` in the cluster's settings, as `llmgw ctl slurm lend`
+and `reclaim` do. Turning Slurm on and choosing its nodes is `llmgw ctl slurm
+setup`, or the Slurm part of the first install. Above the nodes it says where
+the shared files come from and how many users llmgw made; the SHARED column
+is each node's (mounted, exported, missing). `llmgw ctl slurm shared` and
+`llmgw ctl slurm users` change them.
+
 **Logs** shows a node's agent log, newest at the bottom, errors and warnings
 coloured.
 
@@ -203,7 +234,8 @@ and change it with the same commands as `llmgw users`, `llmgw keys` and
 dashboard, and what they belong to. **A** creates an account, already
 approved. On a row: approve one waiting, suspend or reinstate, give or take
 the cluster dashboard, add to or take out of an organization, team or
-project, their limits, a new password (their sessions close), delete, and
+project (on an organization as a member, its manager, who sees its page, or
+its admin, who also changes it), their limits, a new password (their sessions close), delete, and
 **Forget them…** for a request to be forgotten (`llmgw users forget`).
 
 **API keys** — every key, its project, owner, state and last use. **A**
@@ -246,6 +278,28 @@ long a key made there may live, and is the expiry of one made without any
 Limits are requests and tokens a minute, a budget per day or month, and the
 models allowed. Each level only narrows what is above it.
 
+**Sign-in** — how people get in besides a password kept here: single
+sign-on, Active Directory and provisioning, each on or off, with the
+directory's last sync (how many people, accounts kept in step, suspended,
+reinstated).
+- **E** sets up Entra ID. Give the tenant ID, the client ID and secret,
+  and the groups for administrators and operators. A provider other than
+  Entra takes its issuer URL in place of the tenant. The form shows the
+  redirect address to register.
+- **D** sets up Active Directory: the servers, where people are, the
+  service account and its password, the domain's CA certificate as a PEM
+  file on this machine, the group whose members may sign in, and groups
+  for roles and teams.
+- Before it is kept, the directory is tried from the control plane: each
+  server reached and bound, how many people may sign in, how many are in
+  each group. If any step fails, nothing is kept and the steps are shown.
+  **T** tries the settings in force.
+- On the provisioning row, Enter makes a SCIM token, shown once.
+- The password, the client secret and the token are kept in the store,
+  sealed by the master key when one is set. The settings name them as
+  `store:auth/…`, so each control plane has them with no file put on it.
+  The CA is written into the settings themselves.
+
 **Guardrails…** on a row lists the guardrail policies server.yaml defines.
 Each policy is marked in one of three ways:
 
@@ -335,5 +389,18 @@ to restart it.
 
 ## The node's console
 
-On a node, `sudo llmgw` opens `agent.yaml` in the same settings view, and the
-manual. The cluster is watched and changed from a control plane.
+On a node, `sudo llmgw` opens on **Overview**: what this machine is doing, read
+on the machine itself, so it works without an admin token and while the control
+plane is out of reach. Refreshed every three seconds, it shows
+
+- llmgw-agent's state and since when, and the version of llmgw;
+- each control plane the agent dials, from `agent.yaml` and the list the
+  leader last sent, and whether it can be reached;
+- each GPU: how busy, its memory, its heat, and the models on it;
+- the engines running here, from the agent's own records: replica, model,
+  engine, GPUs, port, whether the process is still there, and for how long;
+- the free space where weights and engines are kept;
+- the agent's warnings and errors of the last hour.
+
+Below it, `agent.yaml` in the same settings view, and the manual. The cluster
+is watched and changed from a control plane.

docs/data.md

@@ -72,7 +72,9 @@ each answer that may be reused, as the client got it, and a hash of the
 request it answers, in the gateway process's memory — never in the store,
 never on disk — for the project's TTL or until pushed out by newer ones
 (`router.cache.max_bytes`), and gone with a restart. Never under a `no_store`
-policy. Forgetting a person does not reach into it; the answers age out by
+policy. A project that reuses answers to similar questions (`--similar`)
+also has each question's last user message embedded, and the vector kept
+beside its answer, in the same memory and for as long. Forgetting a person does not reach into it; the answers age out by
 the TTL. The store keeps only the setting, and a usage record's mark that it
 was answered from the cache.
 

docs/guardrails.md

@@ -91,6 +91,21 @@ guard:
   value is the same placeholder throughout a request).
 - **`block`** stops it.
 
+**Putting them back.** With `restore: true` in a policy's `input`, the
+values a request had redacted are put back in its answer: the model reads
+"write to [PERSON_1] at [EMAIL_1]", the client reads the name and address
+it sent. Streamed answers too — a placeholder split across two events is
+held until it is whole. Only that request's own values, only into its
+answer; the map is kept in memory for the request and nowhere else. The
+output rules then read the answer as the client will, so a policy that
+redacts personal data in answers still does.
+
+```yaml
+input:
+  pii: {email: redact, phone: redact}
+  restore: true
+```
+
 **Shapes.**
 
 - `system_prompt` is put first.
@@ -258,6 +273,47 @@ themselves.
     context rather than overflowing it and going unjudged.
 - **Cache:** the same text is not asked about twice within five minutes.
 
+## An outside service
+
+A site required to use a particular guardrail — Azure Content Safety,
+Lakera Guard, NeMo Guardrails, a classifier of its own — names it in a
+policy's `services`. It is asked over HTTP beside the patterns and before
+the judge; its "unsafe" blocks or flags like theirs, and a service that does
+not answer is what the policy's `fail` says.
+
+```yaml
+policies:
+  strict:
+    fail: closed
+    services:
+      - name: lakera
+        url: https://api.lakera.ai/v2/guard
+        header_files: {Authorization: /etc/llmgw/secrets/lakera}   # "Bearer …", read each call
+        body: '{"messages": [{"role": "user", "content": {{json .Text}}}]}'
+        blocked: flagged                 # where the answer says unsafe
+        reason: breakdown.0.detector_type
+        on: both                         # input, output or both
+        action: block                    # or flag
+        timeout: 2s
+```
+
+`body` is a template given `.Text` (`{{json .Text}}` writes it as a JSON
+string; empty sends `{"input": …}`). `blocked` and `reason` are dotted paths
+into the JSON answer, numbers for list items; true, a number above zero, a
+non-empty list or a string other than "false" is unsafe. On input the
+service reads the latest user turn as the model will, with what `redact`
+took out still out; on output the whole answer, once, at its end, as the
+client reads it — with `restore`, the values put back. Being at the end, a
+service can stop a streamed answer only where it is held back
+(`stream: buffer_all`): released a sentence at a time, the client has read
+it by then, and what the service finds ends the stream with
+`content_filter`, flags it and is recorded. The rule recorded is
+`service:<name>:<reason>`. The service is told the text only, never who
+sent it. Under `fail: open`, the default, a service that cannot be asked
+lets the text through — and a caller can make it fail, sending what it
+refuses to read; give `fail: closed` where the service must decide. Of an
+answer asked for with `n` above 1, the service reads the first.
+
 ## Where a policy applies
 
 `guard.default` applies everywhere. Others are attached to a part of the

docs/install.md

@@ -89,13 +89,36 @@ CGO_ENABLED=0 go build -o llmgw ./cmd/llmgw
 
 **The short way is the screens.** On the machine that will be the control
 plane, run `sudo llmgw` with nothing after it. With nothing installed it asks
-whether this is one machine or three for high availability; for three, it asks
-for the other two — address, user, and an ssh key, the ssh agent, or a
-password, and the sudo password if sudo asks for one; or, for a machine
-nothing may log into, a join code to run there with `sudo llmgw join <code>`,
-which dials out to this one on port 7444 — checks all three before
-installing anything (system, room, clock, ports, internet, whether they reach
-one another), shows the plan, and installs. Passwords are held in memory for
+whether this is one machine or three for high availability, then lists what
+an install is made of, each part with what it is for and what is missing
+without it:
+
+| Part | | Its settings (Enter) |
+|---|---|---|
+| Control plane | always | the administrator, other names for the certificate, the portal's language, the sign-up form |
+| Store | always | SQLite or PostgreSQL on this machine; with three machines, PostgreSQL under Patroni on the other two |
+| Control plane on all three | three machines, on | a floating address for whichever leads |
+| Model mirror | on | here, and in which directory, or on another machine |
+| This machine as a node | on for one machine | |
+| Starter model | on for one machine | which: Qwen2.5 0.5B, 1.5B or 3B, Qwen3 4B, 8B or 14B, or any GGUF on Hugging Face; its context |
+| Pictures and 3D | off | |
+| Guardrails' judge | on | refuse or serve flagged when it does not answer; copies; how long it is waited for; what it refuses |
+| Firewall | on | |
+| Notifications | off | the webhooks |
+| Limits and keys | always | requests and tokens a minute; the longest a key lives |
+| Taking turns on a card | always | how long an on-demand model goes unused, and is safe after loading |
+| Offline bundle | off | the bundle and its sha256 |
+
+Space turns a part on or off; Enter opens its settings, which say more. What
+cannot go together is said before anything is done — a starter model with no
+mirror to fetch it into — and closing the message opens the setting that is
+wrong. For three machines it then asks for the other two — address, user, and
+an ssh key, the ssh agent, or a password, and the sudo password if sudo asks
+for one; or, for a machine nothing may log into, a join code to run there
+with `sudo llmgw join <code>`, which dials out to this one on port 7444 —
+checks all three before installing anything (system, room, clock, ports,
+internet, whether they reach one another). Last it shows the plan, part by
+part, what is left out and what that costs, and installs. Passwords are held in memory for
 the run only; a sudo password lets the user sudo without one until the install
 finishes, and is then taken back. At the end it shows the administrator's
 password and the command that does the same from a script.
@@ -129,18 +152,88 @@ That is the whole step. In order, it:
    `nvidia-smi` finds any, on CPU otherwise — and waits for it to install its
    engine
 9. closes llmgw's internal ports to everything but the cluster's networks
-10. fetches Qwen3Guard-Gen 0.6B, the guardrails' judge, and runs two
-    replicas on the CPUs of nodes with room — this one, or the first that
-    join ([guardrails.md](guardrails.md#a-judge-model) has one on a GPU too)
+10. fetches Qwen3Guard-Gen 0.6B, the guardrails' judge, and runs it on
+    CPUs: one replica on up to four cores when this machine is the only
+    node, two on two cores each on the nodes of a larger cluster ([guardrails.md](guardrails.md#a-judge-model) has one on a GPU too)
 11. fetches Qwen2.5 0.5B into the mirror and serves it on this node, so the
     first thing anyone opens `/chat` to find is a model that answers. A
     machine without room for it is told so within two minutes, and why.
 
 `--no-mirror` and `--no-node` leave out 7 and 8, `--no-firewall` 9,
-`--no-guard-model` 10 and `--no-model` 11. An NVIDIA card with no working
+`--no-guard-model` 10 and `--no-model` 11. Every setting the screens ask for
+is a flag too, and the plan shows the command with them: `--starter qwen3-8b`
+(or `hf:<owner>/<repo>/<file>.gguf`) and `--starter-context`, `--mirror-dir`,
+`--judge-fail open`, `--judge-categories`, `--judge-replicas`,
+`--judge-budget`, `--language tr`, `--signup-open`, `--default-rpm`,
+`--default-tpm`, `--max-key-ttl`, `--media` (pictures and 3D: ComfyUI on the
+nodes, and `presets.auto`), `--slurm` (see below), `--evict-after-idle` and `--min-lifetime`. They
+go into `server.yaml` as init writes it, and become the cluster's first
+settings when it starts; a `server.yaml` already there is never overwritten,
+so run again they change nothing — the settings page does. An NVIDIA card with no working
 driver is reported rather than silently ignored; install the driver and run
 init again to serve on it.
 
+**Slurm, for training and batch jobs.** `--slurm`, or the Slurm part on the
+choices screen, runs a Slurm cluster beside llmgw: slurmctld on this
+machine's node, slurmd on every node, those that join later too
+(`slurm.all_nodes`). The agents install munge and Slurm (22.05 or newer)
+from the machines' own package sources, and on this machine slurmdbd, which
+keeps the jobs' records (`sacct`), in a MariaDB installed beside it that
+answers this machine alone. Slurm keeps its records in MySQL or MariaDB
+only, not in the cluster's PostgreSQL; the database's password is made from
+the cluster's munge key, so nothing new is stored. A MariaDB the machine
+already had keeps its settings and only gets the database `slurm_acct_db`
+and its user. Back it up with the machine (`mariadb-dump slurm_acct_db`);
+records older than a year are purged. No node is Slurm's at first: each is
+drained in Slurm and serves models until `llmgw ctl slurm lend NODE` gives it
+to Slurm — its models move off first — and `llmgw ctl slurm reclaim NODE`
+takes it back once its jobs end. Jobs are submitted on this machine with
+`sbatch` and `srun`; `llmgw ctl slurm status` shows where each node is. It
+needs this machine to be a node, and its hostname to be a name Slurm takes.
+On a cluster installed without it, `llmgw ctl slurm setup --controller NODE
+--all-nodes` does the same later. Slurm's ports (6817–6819, and srun's 60001–63000) answer the cluster's
+networks only, as llmgw's own do; a firewall of the site's own between the
+nodes must let them through.
+
+**Slurm's shared files and users.** Jobs need the same files on every
+machine: homes, datasets, output, checkpoints. With `--slurm` this machine
+exports `/shared` over NFS to every Slurm machine and they mount it
+(`slurm.shared`); NFS's port 2049 answers the cluster's networks only, the
+export the Slurm machines alone. That is one machine, not highly available
+and no faster than its disk and network: enough for a few nodes. A site with
+storage of its own uses it instead — `--slurm-shared HOST:/EXPORT` at
+install, or `llmgw ctl slurm shared --path /shared --nfs HOST:/EXPORT`
+later; Lustre, BeeGFS or GPFS the site mounts itself, and `llmgw ctl slurm
+shared --path /lustre/hpc` only has llmgw look that it is there. Every
+active llmgw account is a Linux user on every Slurm machine
+(`slurm.users: llmgw`): its login from its e-mail address, a UID from
+200000 up that is never given to anyone else, its home under
+`/shared/home`. People add their SSH public keys in the portal's
+**Profile** and log in to this machine to run `sbatch`; an account suspended
+or deleted loses its user, its files stay. On the other Slurm machines they
+are let in only while a job of theirs runs there, and their ssh session is
+put in that job, held to its cores and cards (pam_slurm_adopt, which the
+agents install); root and the machines' own users log in as before. A site whose users come from its
+directory (sssd against Active Directory or LDAP) passes `--slurm-users
+site`, or later `llmgw ctl slurm users off`.
+
+Jobs may run in containers: `llmgw ctl slurm containers on` has the agents
+install enroot and pyxis on every Slurm machine (NVIDIA's releases, at
+versions llmgw pins and checks; pyxis is built against each machine's own
+Slurm) and, on GPU machines, nvidia-container-cli from NVIDIA's repository.
+Once every machine has them Slurm loads pyxis, and `ctl slurm status` says
+`containers: on`:
+
+```sh
+srun --gres=gpu:1 --container-image=nvcr.io#nvidia/pytorch:25.01-py3 python train.py
+```
+
+The shared path is in every container where it is on the host, and the job
+runs as its user, held to its cores and cards. Each node pulls the image
+from its registry for each job; for a large image used often, import it
+once (`enroot import -o /shared/images/pytorch.sqsh docker://nvcr.io#...`)
+and give that file to `--container-image`.
+
 **Secure by default.** What init writes is the hardened configuration, not a
 demonstration one:
 
@@ -155,15 +248,43 @@ demonstration one:
 | Requests | every key limited (60 a minute, 200 000 tokens a minute) and priced, so budgets bind |
 | Guardrails | the `baseline` policy on every request: secrets in prompts and answers refused, card numbers masked, injection attempts recorded, and the judge refusing violence, crime, weapons, child abuse, privacy violations, hate, self-harm and sexual content, in prompts and answers |
 
-The judge runs as two replicas, each on a CPU — about half a second for a
-short text on any CPU with AVX2 — and reads a streamed answer every 600
-characters. The policy fails **closed**: a request the judge cannot read
-within 5 seconds is refused with 503 rather than served unjudged, and
-`judge_unavailable` goes to the notification webhooks. Two replicas are what
-make that bearable; on one small machine, or a VM without AVX2 (see
+The judge runs on CPUs. A judgment of about 280 tokens took 2.3 s on four
+cores, 4.2 s on two and 8.0 s on one (a 4-core VM). It also reads a
+streamed answer every 600 characters. The policy fails **closed**: a request
+the judge cannot read within 5 seconds is refused with 503 rather than
+served unjudged, and `judge_unavailable` goes to the notification webhooks.
+So a judge gets cores of its own: on one machine one replica on up to four,
+and on a larger cluster two replicas of two cores each, spread over its nodes
+so that losing one leaves the other. On a machine of fewer than four cores,
+a busy one, or a VM without AVX2 (see
 [step 6](#6-setting-the-nodes-cpu-model-if-it-is-a-vm)), the judge is the
-first thing to fail, so `fail: open` — serve it, recorded as flagged — or
-`--no-guard-model` is the choice there. Defamation, specialized advice,
+first thing to fail. The choices there are a longer `latency_budget`,
+`fail: open` (serve the request, recorded as flagged), the judge on a GPU,
+or `--no-guard-model`.
+
+init times the judge once it is up. If a judgment takes more than half the
+wait, or more than the wait when several come at once, it says so before the
+next steps, with what to do. The wait itself is not changed without being
+asked. `llmgw ctl guard judge check` does the same at any time:
+
+```
+qwen3guard-0.6b, 2 replica(s) ready, a request waits 5.0 s for it
+  alone:    2.4, 2.3, 2.3 s
+  4 at once: slowest 6.1 s; about 0.7 judgments a second, so about 0.3 requests a second with short answers
+tight: a busy moment passes the budget
+```
+
+In a three-machine install the judges are better on machines of their own.
+On a 4-core control plane that was also a node, 38 of 60 requests under load
+were refused for want of a judgment. With the judges on a node of their own
+(4 cores, two judges of two cores each), 2 were refused. The setup screen asks
+for up to two **judge machines** after the store's two, or give
+`--judge-node user@host` to `server init`, once for each. Each is added as a
+node labelled `llmgw.role: judge`, and the judge's profile selects that
+label, so the judges run only there and nothing else does. Four cores and
+8 GB are enough for two judges.
+
+Defamation, specialized advice,
 intellectual property and elections are left out of what it refuses: they
 catch too much ordinary work. [guardrails.md](guardrails.md) has the rest.
 
@@ -331,7 +452,11 @@ by itself. Three things are worth knowing before you edit it:
 - `default_rpm`, `default_tpm` and `default_price` are what make limits and
   budgets mean anything. Without a price every request costs zero and a budget
   never binds; without the defaults, a project created without limits is
-  metered by nothing at all.
+  metered by nothing at all. A price may set `cached_per_1k` beside
+  `prompt_per_1k`: the prompt tokens the engine or provider read from its
+  cache (the answer's `cached_tokens`) are charged at it, the rest at the
+  prompt rate. A provider that bills cached reads at a tenth is
+  `{prompt_per_1k: 2500, cached_per_1k: 250, completion_per_1k: 10000}`.
 - `control.enabled: true` is what makes this a control plane while its catalog
   is empty. Without it, a file with no models is refused as having nothing to
   serve.
@@ -639,6 +764,8 @@ Nothing to do: the agent installs them once the node joins, under
 | vLLM | nodes with an NVIDIA GPU it can run on | a Python environment built by uv from PyPI, about 8 GB |
 | SGLang | only when you ask for it | the same way as vLLM |
 | TensorRT-LLM | only nodes where you installed it | not installed by llmgw: see [TensorRT-LLM, which you bring](#tensorrt-llm-which-you-bring) |
+| FreeToken | nodes with a GPU of compute capability 8.0 or newer (RTX 30, A100 and later), driver 580+, x86-64 and the CUDA 13 toolkit | a Python environment built by uv from PyPI (`freetoken[accel]`); see [FreeToken](#freetoken-models-larger-than-the-card) |
+| Strata | only when you ask for it (`engines: {strata: "0.2.7"}`) | its source, pinned by digest, in a Python environment, and its prebuilt engine for CUDA 13 (driver 580+) or CUDA 12.9 (Pascal to Ada, driver 575+); see [Strata](#strata-a-large-model-of-experts-on-one-card) |
 
 vLLM is resolved as PyPI stood a day after that vLLM was published, so every
 node gets the same dependencies whenever it is built, and installed with
@@ -681,6 +808,100 @@ you installed yourself on a node, name its path in that node's `agent.yaml`
 (`engines.llamacpp.bin`, `engines.vllm.bin`, `engines.sglang.python`,
 `engines.trtllm.python`): the agent then installs none of its own beside it.
 
+#### Strata: a large model of experts on one card
+
+Strata runs a few mixture-of-experts models — Qwen3.8-Flash-Next (125B) and its
+Swift, Coder, Unsloth, Gyro and NVFP4 versions, Qwen3.6-35B-A3B, DeepSeek
+V4.1-Flash — with the experts in the node's RAM and the rest on one card of 12
+GB or more. Where it is installed, the plan offers it for a repository of one
+of those families:
+
+```
+llmgw ctl model plan ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF
+llmgw ctl model serve ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF \
+  --file Qwen3.8-Flash-Next-GSQ-RCO-IQ2_XS-00001-of-00002.gguf --engine strata
+```
+
+The Hugging Face explorer shows the same ("1× … with Strata, N GB of RAM"),
+and the mirror's serve form serves such a version on Strata.
+
+The mirror page's **Strata** tab lists every family and size Strata runs, with
+what each downloads and keeps in RAM and where it would run here:
+
+- **Fetch into the mirror**, for the GGUF families: the size's files, checked
+  against Hugging Face's checksums; the version's serve form then serves it on
+  Strata.
+- **Serve**, for the packs Strata fetches itself — the NVFP4 packs
+  (`Maximilian228/Qwen3.8-Flash-Next-NVFP4-GPTQ-Strata` and its NVIDIA
+  sibling, NVIDIA's FP4-Local) and DeepSeek V4.1: the model is added with the
+  repository as its source, and its first start fetches tens of gigabytes onto
+  the node, each file pinned and checked by Strata, before preparing it.
+  `ctl model serve` does the same for these repositories.
+
+Repositories of architecture `strata-ple` that Strata 0.2.7 does not list —
+community packs such as `…-Q8_0-Strata` — are named as such in the explorer:
+Strata's setup refuses them, and llama.cpp cannot read them. A Strata profile
+says its family and size (`--family qwen --size IQ2_XS`), which the catalog
+checks against what Strata runs, and reserves the RAM the size keeps its experts
+in (48 GB and the engine's 4 for IQ2_XS) unless it says `min_memory_gb`. Its
+first start prepares the model on the node, which takes minutes, and fetches a
+draft layer of about 5 GB from the internet.
+
+#### FreeToken: models larger than the card
+
+[FreeToken](https://github.com/FlashML-org/FreeToken) serves mixture-of-experts
+models several times a card's size from one card: the experts stay in the
+node's RAM, the ones used most are cached in VRAM, and the rest stream over
+PCIe. Unlike Strata it reads Hugging Face safetensors checkpoints as they are
+published — DeepSeek V4 Flash, GLM-5.x, Qwen3.6/3.8 (Flash-Next among them),
+gpt-oss, Gemma 4, MiniMax M2.5/M3 — in FP8, NVFP4, MXFP4 or BF16, and batches
+several requests (`--max-running-requests`, 4 by default).
+
+Every node that can run it installs it, as with vLLM: an NVIDIA GPU of compute
+capability 8.0 or newer, driver 580 or newer, x86-64. Older cards are passed
+over without a word — on a Quadro RTX 6000 (7.5) version 0.2.1 had no kernel
+image for the card's first prefill, and in fp16 its first request ended in an
+illegal memory access. Its FP8 checkpoints need 8.9 (Ada, Hopper) and its NVFP4
+ones 12.0 (RTX 50, RTX PRO 6000 Blackwell). It compiles some kernels the first
+time a model loads, so the node also needs the **CUDA 13 toolkit** (`nvcc` in
+`CUDA_HOME`, `/usr/local/cuda` or `/usr/local/cuda-13.*`); a node of new cards
+without one says so on the cluster page instead of installing an engine that
+fails on its first model.
+
+A FreeToken profile runs on one card of one node (`llmgw ctl model add -f`):
+
+```yaml
+name: glm-5.2
+source: /var/lib/llmgw-weights/GLM-5.2-NVFP4   # a safetensors checkpoint's directory
+profiles:
+  - name: one-card
+    engine: freetoken
+    gpu_class: rtx-pro-6000-96g
+    min_memory_gb: 200            # the experts live in RAM: say how much
+    args: ["--max-running-requests", "8"]
+```
+
+A Hugging Face id as the source works too — FreeToken downloads the
+repository itself — but it is not pinned to a commit or checked against
+Hugging Face's checksums as vLLM's are, FreeToken having no `--revision`; a
+directory the mirror fetched, or one put on the node, is.
+
+The agent gives it the model, its address, its card (`--gpu`) and the key (in
+`FREETOKEN_API_KEY`, not on the command line); `--mmproj`, media paths and
+domains, `--hf-overrides` and `--moe-cache-split` are refused as they read
+files or fetch. `ctl model context` sets `--max-seq-len-override`. Readiness
+waits for its `/health` to stop saying `loading`, since it answers 200 from the
+moment its HTTP server is up. It serves no Prometheus metrics, so the router
+counts its requests itself.
+
+llmgw follows FreeToken release by release: the version is
+`DefaultEngines["freetoken"]`, and a weekly workflow
+(`.gitea/workflows/engine-releases.yml`) fails when PyPI has a newer one, the
+notice to check it and move to it. A version of your own is named under
+`engines` like any other (`freetoken: "0.2.2"`), and `freetoken: ""` installs
+none; one node is turned off with its build set to *none* (console, the node's
+menu).
+
 #### Choosing a node's vLLM build
 
 Which build a GPU node installs is chosen on the cluster page, in the **vLLM
@@ -812,6 +1033,45 @@ Nothing to do here if you built the mirror in step 3 — the agent fetches what
 model's `source` names, and that source is your mirror. This section is for the
 arrangements where the files are already on the machine.
 
+**Weights in a pickle format are refused.** `.bin`, `.pt`, `.pth`, `.ckpt` and
+`.pkl` are Python pickles, and loading one runs whatever it holds, as the
+engine's user, on the node — which is how the malicious models found on Hugging
+Face work. A model whose source is one, or a mirror version holding one, is
+refused when it is added and by the node before it fetches anything; vLLM and
+SGLang given a model by its Hugging Face id read only its safetensors. Take the
+repository's safetensors or GGUF. A model that has no other form, and whose
+files were checked, can say so:
+
+```yaml
+- name: triposr
+  source: http://mirror:8081/triposr/<rev>/model.ckpt
+  allow_pickle: true
+```
+
+A local path is judged by its name as any source is; a local directory is
+not looked into. What an engine downloads itself by a Hugging Face id, past
+the node — an adapter given by its id, a TensorRT-LLM model's — is refused
+when the model is added, unless it says `allow_pickle`: mirror its
+safetensors and give that instead.
+
+**Weights are pinned to a commit and checked.** A model added with
+`ctl model add`, the API or the console, whose weights come from Hugging Face,
+is pinned to the commit the repository is at when it is added:
+
+- `hf:org/repo/file` and a `…/resolve/main/…` URL become that commit's URL;
+- a repository vLLM or SGLang downloads itself is given `--revision`;
+- the sha256 Hugging Face published for each file goes in the model's origin
+  (its definition, `GET /api/v1/models/NAME`, shows it).
+
+A node then refuses a copy of the file that is not it, naming both checksums,
+and deletes it; a cached copy is checked once and its checksum kept beside it.
+The mirror does the same for a model `ctl model serve` fetches, and
+`llmgw mirror add` for a Hugging Face URL (`--sha256` for any other). A new
+upload to the repository reaches the cluster only through `ctl model update`,
+which says what changed. Not pinned: models declared in the cluster's settings,
+which are not added this way, and a vLLM model whose arguments already say
+`--revision` keeps that one.
+
 Put them where the node can read them. `scp`, a shared filesystem, or the
 engine's own downloader are all valid, and which you want depends on how many
 nodes you have.
@@ -970,6 +1230,27 @@ A node with no location is placed as before, as a rack of its own.
 See [facility-aware-placement.md](facility-aware-placement.md) for power caps
 and holds.
 
+### A card in the wrong slot
+
+An x16 card in a slot wired x8 or x4, behind a riser, or in a slot of an
+older PCIe generation than the card's, links slower than it was built to and
+works otherwise. A model on the card still decodes at the card's own memory
+speed, but it loads slower, a replica split over several cards by PCIe is
+slower, and so is anything that spills to the host. Each card's agent reads
+how its link stands — the fastest generation card and slot allow together,
+and the lanes in use — against what the card could do, and the node says so
+among its warnings (the dashboard, the console's Nodes, `llmgw doctor`):
+
+```
+GPU 1 (NVIDIA H100 80GB HBM3) linked at x8 of its x16 lanes: 31.5 GB/s where it
+could have 63.0. Check the slot (wired fewer lanes, or an older generation), a
+riser, or the BIOS's PCIe setting
+```
+
+The generation a card reports now drops while it idles, to save power; what is
+compared is the fastest it may reach, which does not. `lspci -vv -s BUS-ID`
+shows the same: `LnkCap` is the card's, `LnkSta` the link's.
+
 ## 6. Setting the node's CPU model, if it is a VM
 
 Proxmox defaults a VM to `kvm64`, which hides AVX, AVX2, FMA and even SSE4.2
@@ -1287,6 +1568,26 @@ catalog, and a key rotated in the file is used within seconds. A model whose key
 file cannot be read is not offered. Everything else applies as to any model:
 prices and budgets, rate limits, usage records, and key allowlists.
 
+A provider's limits are per key. With several keys, one answering 429 need not
+hold up the model: give them all, each with an optional weight (its share
+against the others'), in place of `api_key_file`:
+
+```yaml
+  external:
+    url: https://api.openai.com
+    model: gpt-4o-mini
+    api_key_files:
+      - {file: /etc/llmgw/secrets/openai-1.key, weight: 2}
+      - {file: /etc/llmgw/secrets/openai-2.key}
+```
+
+Each key is an endpoint of its own, `external:cloud-mini#1`, `#2`. A key that
+answers 429 is passed over while its breaker is open, and the request goes to
+another. One that answers 401 (revoked, expired, mistyped) is passed over too,
+and `external_key_refused` goes to the notification webhooks, naming the
+endpoint and never the key. The usage records say which key answered by the
+same name. A key whose file cannot be read is left out, and the others stay.
+
 **Its prompts leave the cluster.** Decide who may send them: a key or project
 whose allowlist leaves the external model out cannot reach it, as a fallback or
 by name. The cluster page and `ctl models` mark it `external`.
@@ -1503,6 +1804,93 @@ Not yet: models declared in server.yaml cannot name a peer's model as a
 fallback (the catalog in the file is checked before the peers are read), and
 a peer behind a private CA needs that CA in the system's trust store.
 
+### Stopping a model, and removing one with its files
+
+```
+llmgw ctl model stop qwen-7b             # none of its replicas run; it is kept
+llmgw ctl model start qwen-7b            # runs again, scaled as it was
+llmgw ctl model remove qwen-7b           # leaves the catalog; its files stay
+llmgw ctl model remove qwen-7b --purge   # its files on the mirror and the nodes too
+```
+
+- **Stop** keeps the model in the catalog with its files and what it was
+  scaled to; a request for it is told an administrator stopped it. `ctl
+  models` shows it as stopped. A `ctl scale` while it is stopped starts it at
+  that number.
+- **Remove** leaves its files on the mirror and the nodes, for adding it
+  again without fetching it. `ctl weights` lists such files on the nodes.
+- **Remove --purge** deletes every version of it on the mirror (a second
+  mirror follows), its adapters and the weights its last update replaced,
+  and the nodes' copies once its replicas have stopped. What another model
+  uses — a canary's weights, a shared adapter — stays. It cannot be undone:
+  adding it again fetches it again.
+
+The console has the same in a model's menu: Stop… or Start, Remove…, and
+Remove with its files….
+
+### Changing a served model's context and longest answer
+
+How much one request may use, and how long an answer may be, are set when a
+model is served. To change them afterwards, in place:
+
+```
+llmgw ctl model context qwen-7b 32768                     # tokens a request
+llmgw ctl model context qwen-7b 32768 --slots 4           # llama.cpp: four at once
+llmgw ctl model context qwen-7b 32768 --max-output 8192   # and the longest answer
+llmgw ctl model context qwen-7b --max-output 4096         # the longest answer alone
+```
+
+The longest answer is the gateway's: a request asking for more `max_tokens`
+is refused, `/v1/models` lists it as `max_output_tokens` for clients to read,
+and the account page's chat asks for no more. Changed alone it is only
+the catalog's, in force at once, and no replica is started again. With a
+context it is set with it; left out, it grows and shrinks with the context
+as before, unless it was set to something else. It never exceeds the context.
+
+For a context, the catalog, the engine's arguments (`--ctx-size` and
+`--parallel` for llama.cpp, `--max-model-len` for vLLM, `--context-length` for
+SGLang, `--context` for Strata's setup, `--max-seq-len-override` for
+FreeToken) and the memory its KV cache takes change together. Strata and
+FreeToken size their caches to the card themselves, so their cards are not
+checked here. A context longer than the
+model was trained for is refused, and so is one its cards cannot hold — with
+the figures, so you know how far to go. Its replicas are then started again on
+it, one by one, a new one ready before the old one stops. The console has the
+same under a model's menu, Context length…, with the longest answer beside
+it. The admin API is `POST /api/v1/models/{name}/context` with `tokens`,
+`slots` and `max_output`, any of them left out. A model declared in the cluster's
+settings is changed there (`llmgw ctl config edit`).
+
+### Moving a served model to more cards
+
+How many cards of one node each replica uses can be changed while the model
+is served:
+
+```
+llmgw ctl model gpus qwen3-coder-30b 2              # two cards a replica
+llmgw ctl model gpus llama-70b 4 --profile h100     # the model has several profiles on cards
+llmgw ctl model gpus qwen3-coder-30b 2 --in-place   # no room beside the running replica
+```
+
+vLLM, SGLang and TensorRT-LLM spread the model by tensor parallel (`tp`),
+which needs the count to divide the model's KV heads or be divided by them:
+a model with 8 splits 2, 4, 8 or 16 ways, not 3 or 12. llama.cpp has no
+tensor parallel; it splits the layers over the cards (`gpus_per_node`), and
+a `--tensor-split` written for the old cards is dropped. ComfyUI and Strata
+run on one card. A count no node has is refused, and so is one whose cards
+would not hold the weights and the context — with the figures; fewer cards
+with a long context may need `llmgw ctl model context` first.
+
+Its replicas are then started again on it, each new one ready beside the old
+before the old stops. That needs the cards free beside it: on a node of two
+cards with the model on one, two do not fit until it is gone, and the command
+says so. `--in-place` stops each replica first and starts its replacement on
+the cards it held and the free ones, and the model does not answer while the
+new one loads — a model with several replicas keeps answering on the others.
+A replica spread over several nodes is laid out by hand, with a fabric
+(`nodes`, `gpus_per_node` in the profile). The console has the same under a
+model's menu, GPUs….
+
 ### Trying new weights on some of the traffic first
 
 When a model's repository publishes new weights, `llmgw ctl model update`
@@ -1762,12 +2150,31 @@ dashboard, is in [deploy/monitoring/](../deploy/monitoring/) — including where
 to run it, which is not the control plane: monitoring that dies with the thing
 it monitors tells you nothing at the moment you need it.
 
-Two figures repay knowing about. `llmgw_gpu_utilization_percent` is absent
+Three figures repay knowing about. `llmgw_gpu_utilization_percent` is absent
 rather than zero when the driver does not report it — a flat zero across a
 fleet means the measurement is missing, not that the fleet is idle. And
 `llmgw_gpu_nvlink_peers` at zero everywhere means the topology is unknown,
 which is what a binary built without the nvml tag reports.
 
+The third is what "busy" means. `llmgw_gpu_utilization_percent` is
+nvidia-smi's `utilization.gpu`: the share of time *any* kernel ran, however
+little of the chip it ran on — a kernel on one SM of 142 reads 100%. A Quadro
+RTX 6000 decoding read 92% busy while its SMs had work 69% of the time and
+issued an instruction on 18% of their cycles. How much of the chip is at work
+is in `llmgw_gpu_sm_active_percent`, `llmgw_gpu_sm_occupancy_percent`,
+`llmgw_gpu_tensor_active_percent` and `llmgw_gpu_dram_active_percent`, from
+the card's profiling counters. The agent reads them from NVML's GPM on
+Hopper and later, with nothing to install; on older cards it reads them from
+a [dcgm-exporter](https://github.com/NVIDIA/dcgm-exporter) on the node, at
+`http://127.0.0.1:9400/metrics` unless `discovery.dcgm_exporter` in
+agent.yaml says otherwise (`off` reads none). Where neither is there, the
+dashboard's GPU map and the console colour cards by their draw against the
+power limit (`llmgw_gpu_power_watts / llmgw_gpu_power_limit_watts`), which
+rises with how much of the chip works rather than with whether anything
+does. Low tensor-core activity in decode is not waste: a quantised
+matrix-vector product at batch one runs on the integer units, and it is
+prefill and batches that use the tensor cores.
+
 ### Token by token, and SLOs
 
 Beside time to the first token, every streamed answer is timed per token:
@@ -1877,6 +2284,18 @@ So "why did qwen take four minutes to start" is one trace: find `replica start`
 with `llmgw.model` = qwen, and the node's spans beneath it say whether the
 time went to the download, the launch or the engine loading.
 
+An inference request and each `upstream` attempt also carry the
+[OpenTelemetry GenAI conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/),
+so a tool that reads them — Grafana, Jaeger, Langfuse, Arize Phoenix — shows
+llmgw's calls as model calls without a dashboard made for llmgw:
+`gen_ai.operation.name` (`chat`, `text_completion`, `embeddings`),
+`gen_ai.request.model`, `gen_ai.request.max_tokens`, `temperature` and `top_p`
+when the client set them, `gen_ai.conversation.id` (the session),
+`gen_ai.response.model`, and `gen_ai.usage.input_tokens`, `output_tokens` and
+`cache_read.input_tokens`. On an `upstream` span `gen_ai.provider.name` is the
+engine that answered (`vllm`, `llamacpp`, `external`, …). What was asked and
+what was answered are never in a span.
+
 A node follows the control plane's decision: a start the control plane kept is
 traced on the node whatever the agent's own `sample_ratio`, and one it sampled
 out is not; the agent's ratio decides only for what arrives with no trace, as
@@ -2041,8 +2460,10 @@ llmgw keys cache --project Acme/ml/train --off
 - **Policies still apply:** the limits are checked and the guardrails judge
   the request before the cache is looked in, and the output policy judges a
   reused answer as it would a new one. A request continuing a stored
-  conversation, or under a `no_store` policy, is never answered from the
-  cache nor kept in it.
+  conversation, under a `no_store` policy, or whose redacted values are put
+  back in its answer (`restore: true`) is never answered from the cache nor
+  kept in it: redacted, two callers' requests read the same, and the answer
+  has the first caller's values in it.
 - **What a hit costs:** nothing. It is recorded as a request — status 200,
   the model, the key, the rate limit's count — with no tokens and no cost,
   marked as a cache hit (`llmgw ctl request <id>` says so), so it spends no
@@ -2062,8 +2483,46 @@ router:
     max_entry_bytes: 2097152 # 2 MB
 ```
 
-`llmgw_cache_requests_total{model,result}` counts hits, misses and
-bypasses, `llmgw_cache_bytes` and `llmgw_cache_entries` what is held. The
+### Similar questions
+
+A help desk asked "how do I reset my password" in forty wordings asks the
+model forty times under the exact cache. `--similar` also gives a chat
+question the answer to one like it:
+
+```
+llmgw keys cache --project Acme/support/desk --ttl 24h --similar 0.95
+```
+
+```yaml
+router:
+  cache:
+    embedding_model: bge-m3   # an embedding model in the catalog
+```
+
+- **What is compared:** the last user message, embedded by
+  `router.cache.embedding_model`, against the questions already answered
+  in the same conversation — the same system prompt, the same earlier turns,
+  tools, `response_format` and sampling settings, the same model and
+  project. A like question in another conversation, after a tool's result,
+  or with a picture, a sound or a file in it, is not taken for this one. The exact cache is asked first, and an
+  exact hit embeds nothing.
+- **How alike:** the cosine similarity of the two embeddings, at least the
+  project's `--similar`. 0.95 is strict; below 0.9 different questions start
+  to meet. A question that differs only in a number or a name ("2+2" and
+  "2+3") can be closer than that, so a project whose questions turn on
+  figures should not use it. A question over 2,000 characters is reused only
+  when it is the same.
+- **Said:** `x-llmgw-cache: similar`, with `x-llmgw-cache-similarity`
+  and `Age`. Recorded and charged as a hit. The same rules as a hit decide
+  which answers are reused and kept; guardrails judge the request before.
+- **The embedding** is asked of the model by the gateway on its own account,
+  as a judge is, within 2 seconds; every miss waits for it. Without the
+  model, or when it does not answer, only the same request is reused. The
+  vectors are kept in memory with their answers and counted in
+  `router.cache.max_bytes`.
+
+`llmgw_cache_requests_total{model,result}` counts hits, similar hits, misses
+and bypasses, `llmgw_cache_bytes` and `llmgw_cache_entries` what is held. The
 console has it as **Cache answers…** on a project.
 
 ## Batches
@@ -2172,6 +2631,7 @@ triposr`. What that adds is this entry, and its workflow is built in:
 ```yaml
 - name: triposr
   source: hf:stabilityai/TripoSR/model.ckpt
+  allow_pickle: true               # published as this checkpoint only
   comfyui:
     workflow: triposr              # built in; a model of your own names a JSON file on every control plane
     output: meshes
@@ -2319,6 +2779,62 @@ Without authentication, `threads.store` and `threads.retain_hours` in
 `server.yaml` keep both conversations and notes in one shared pool. That is for
 one trusted machine only; see [development.md](development.md).
 
+## Three machines, each a control plane
+
+The screens' three machines are high availability for the gateway as well as
+the store: the two database machines are made control planes too, standing
+by while the first leads — what `server add`, below, does for one machine.
+By command, `--standbys` with `--patroni`, and `--vip` for the floating
+address:
+
+```
+sudo --preserve-env=SSH_AUTH_SOCK llmgw server init --admin you@example.com \
+    --patroni ubuntu@10.0.0.6 --patroni ubuntu@10.0.0.7 \
+    --standbys --vip 10.0.0.160/24 --no-node --no-model
+```
+
+| Machine | Runs |
+|---|---|
+| the first | control plane, leading at first; model mirror; etcd |
+| the other two | control plane, standing by; PostgreSQL under Patroni; etcd |
+
+etcd votes on all three and PostgreSQL is on two, so any one machine can be
+lost: a standby leads within seconds, Patroni promotes the standby database
+if it was the primary, and the floating address follows whichever leads. The
+mirror's management API answers the three control planes' addresses, not
+only its own machine's, so a standby that comes to lead can still fetch and
+add models.
+
+**A second mirror.** The mirror is on the first machine, and without a
+second one, models the nodes already have keep serving while that machine
+is down, and new ones wait for it. With a second, the cluster carries on.
+- **Setting it up:** the screens offer one on either database machine
+  ("A second mirror"); by command, `--second-mirror 1` (or `2`). For an
+  install that has one mirror, run this on the mirror's machine:
+
+  ```
+  sudo llmgw mirror add-peer ubuntu@10.0.0.7 --control-plane 10.0.0.5 --control-plane 10.0.0.6 --control-plane 10.0.0.7
+  ```
+
+  It logs in with root's ssh key on the mirror's machine, so that key must
+  be accepted for the user named on the second machine. When the mirror has
+  a machine of its own, without the control plane, the command cannot change
+  the cluster's settings itself: it prints the `llmgw ctl config set
+  mirror.peers …` to run on a control plane.
+- **Kept in step:** each mirror reads the other every thirty seconds and
+  copies what it lacks, keeping a copy only if its checksum matches.
+  - A model removed from one is removed from the other, even if that one
+    was down at the time and comes back later.
+  - A model added again after it was removed stands on both.
+  - Neither mirror leads.
+- **Control planes:** they name both (`mirror.peers` in the cluster's
+  settings) and send the mirror's management calls to the first that
+  answers.
+- **Nodes:** while the mirror a model's address names is down, a node is
+  given the same file on the other.
+- **Disk:** the second machine needs room for every model, as the first
+  does.
+
 ## Two control planes
 
 Once the store is replicated (`server init --patroni`, above), one command on
@@ -2335,8 +2851,12 @@ and lets it into the store; it starts there as a standby. Every node is told
 of it within half a minute and fails over to it without being touched.
 `--vip` keeps a floating address on whichever control plane leads, with
 keepalived on both: that is the address to give clients, or a proxy in front.
-Without it, put a load balancer in front that checks `/ready` — not
-`/health`, which is 200 on a standby that would refuse every request.
+Without it, put a proxy in front that sends to whichever answers `/leader`
+with 200 — not `/health`, which is 200 on a standby too, and not `/ready`,
+which is 503 on a leader with no model yet and would take the portal away
+with it. The end of [deploy/proxy/Caddyfile](../deploy/proxy/Caddyfile) has
+Caddy doing that for three control planes, and
+[deploy/ha/](../deploy/ha/README.md#check-leader-not-health) why.
 
 - **What a failover costs.** Measured on the reference cluster, stopping the
   leader: the standby led within five seconds, every node had moved to it
@@ -2362,9 +2882,16 @@ Without it, put a load balancer in front that checks `/ready` — not
   `server add` checks, and says what to add to the mirror's unit if it
   refuses the new control plane — which would otherwise find out the first
   time it leads and fetches a model.
-- **Limits are per process.** Rate limits are counted in memory, and clients
-  reach the leader through the floating address, so a failover starts each
-  key's minute afresh.
+- **Standbys can answer clients too.** With `router.active_active: true`
+  (`llmgw ctl config set router.active_active true`) a standby answers `/v1`
+  itself, routing to the leader's replicas, instead of passing it on: more
+  machines carry the traffic. Point clients at every control plane — DNS,
+  or your own load balancer checking `/ready` — since the floating address
+  stays on the leader. Rate limits stay the cluster's: each router's counts
+  reach the others every two seconds, and a standby that becomes the leader
+  keeps the minute it was counting. Without the setting, rate limits are
+  counted in the leader's memory and a failover starts each key's minute
+  afresh.
 
 [deploy/ha/](../deploy/ha/) describes what each half does, for anyone putting
 their own load balancer in front.
@@ -2397,6 +2924,89 @@ the same files and keys as `locales/en/`, and its `_meta.*` entries saying how
 it writes numbers and dates. The tests refuse a catalogue that lacks a key a
 page uses.
 
+## The company's MCP servers
+
+llmgw's `/mcp` can serve the company's own MCP servers beside its own tools.
+Each server, a ticket system, a file server or a database, is registered once.
+Its credentials stay on the control planes, and people's agents are given only
+`/mcp` and their llmgw key.
+
+```
+llmgw ctl mcp add jira --url https://jira.internal/mcp \
+    --description "Acme Jira" --secret-file Authorization=/root/jira-token
+llmgw ctl mcp test jira          # its tools, asked with its credentials
+llmgw keys limits --scope team --id Acme/support --tools 'jira.*'
+```
+
+- **Credentials:**
+  - `--secret-file NAME=FILE` reads a header's value once and keeps it in
+    the store, sealed by the master key, so no file has to be on each
+    control plane.
+  - `--header-ref NAME=REF` reads it at each call from a file, `env:`,
+    `vault:` or `store:` instead.
+  - `--header NAME=VALUE` is for what is not secret.
+- **Who may call what:** tools are named `server.tool` and granted as
+  patterns (`jira.*`, `db.query`) at any level of the tenancy tree, as
+  models are.
+  - A level that names any narrows what is above it: a tool is seen where
+    every level above that names tools names it too. A team under an
+    organization that names `hr.*` sees no `jira.*` whatever the team is
+    given, until the organization names it as well.
+  - Unlike models, **none granted anywhere is none**: a tool reaches into a
+    company system, so it is granted, never assumed.
+  - `keys limits --tools` writes a level's list whole, every server's
+    together; an empty list (`--tools ''`) closes the level to every tool.
+    The console's **MCP servers** screen and the `/admin/mcp` page grant
+    and take back one server's tools: a level given its first list starts
+    from the nearest list above it, so it keeps what it saw; one left
+    naming none goes back to naming none, or stays closed when a level
+    above names tools, so taking one server away opens it to no other. They
+    say when a level above keeps the granted tools out of reach.
+- **The server is told who calls:** `X-Llmgw-Key`, and `X-Llmgw-User` and
+  `X-Llmgw-Project` where there are some. It is called with llmgw's
+  credentials, never the caller's key.
+- **Bounds:** a call has 60 seconds (`--timeout`) and an answer 4 MB. A
+  server's progress is passed on to a client that takes a stream. `--rate N`
+  allows each caller N calls a minute to a server's tools.
+- **Guardrails, both ways:** the caller's policies read a tool's answer as
+  the prompt it becomes.
+  - An instruction planted in a ticket is caught as a pasted one would be
+    (`injection: block`).
+  - Personal data is redacted or refused as the policy says. A refused
+    answer does not reach the agent.
+  - The arguments are read before they leave, so a secret pasted into one
+    does not reach the company system.
+- **A server that keeps failing:** five calls failing in a row take a server
+  out for a minute. Its tools leave the list, and a call is answered at once.
+- **Record:** every call is in `llmgw ctl audit`, with its arguments' size
+  and a digest, never their text. `llmgw_mcp_tool_calls_total` and
+  `llmgw_mcp_tool_call_seconds` count them by server, tool and outcome.
+- **Changes:** a server added, changed or removed on one control plane is
+  served by every one within half a minute; its tool list is read again
+  every five minutes. `--disabled` keeps a server but serves none of its
+  tools.
+
+**This cluster's own, read only.** `llmgw ctl mcp add llmgw --url
+llmgw:admin` registers the cluster itself as a server: tools that say what
+`llmgw ctl` does, and change nothing. It is answered in the control plane's
+own process, with no address and no token to give.
+
+| Tool | Says |
+|---|---|
+| `llmgw.cluster_status` | control planes, store, nodes, models against what they asked for, GPUs, what is unhealthy |
+| `llmgw.list_nodes` | each node's state, GPUs, load, fabrics, agent version and engines |
+| `llmgw.list_models` | each model's replicas ready against wanted, profiles, why one is parked |
+| `llmgw.list_replicas` | each replica's shards: node, GPUs, state (`model` for one model's) |
+| `llmgw.explain_placement` | why a model is not placed, node by node |
+| `llmgw.node_logs` | a node's last agent log lines, or an engine's (`source`) |
+
+Grant it as any other: `llmgw keys limits --scope key --id KEY-ID --tools
+'llmgw.*'`. It is the same as an administrator's view, so grant it to the
+people who would have `ctl`.
+
+Servers over HTTP only for now (the streamable HTTP transport). A server
+that runs as a local process (stdio) is planned.
+
 ## Single sign-on
 
 People can sign in through your identity provider — Keycloak, Entra ID,
@@ -2404,8 +3014,10 @@ Google, Okta, Authentik, anything that speaks OpenID Connect.
 
 1. Register a client there with the redirect address
    `https://<gateway>/login/oidc/callback`.
-2. Put its secret in a file only root reads.
-3. Add to `server.yaml`:
+2. Put its secret in a file the llmgw user reads, on every control plane.
+3. Add this to the cluster's settings with `llmgw ctl config edit`. On a
+   cluster with a store, `auth` is the cluster's: the same block in
+   `server.yaml` is not read, and the log says so.
 
 ```yaml
 auth:
@@ -2454,6 +3066,99 @@ and point it at `https://<gateway>/scim/v2` with that token as the bearer.
   join and leave them. Membership in a group is removed when someone leaves
   it.
 
+### Entra ID
+
+The console sets this up: `sudo llmgw`, People → Sign-in, **E**. What
+follows is what it writes, and what to do in Entra.
+
+Entra ID is an OpenID Connect provider like the others, with two things to
+set in the app registration:
+
+1. **Redirect URI:** under Authentication, add a Web platform with
+   `https://<gateway>/login/oidc/callback`. The issuer is
+   `https://login.microsoftonline.com/<tenant-id>/v2.0`.
+2. **Groups in the token:** under Token configuration, add a groups claim.
+   Choose **Groups assigned to the application**, and assign the groups you
+   map in `roles` and `memberships` to the enterprise application. Entra
+   sends groups by object id unless you choose to emit names
+   (`sAMAccountName` for groups synced from AD), so write them the same way
+   in the mapping.
+
+Entra leaves the groups out of the token for someone in more than 200 of
+them and sends a pointer to Graph instead. llmgw then keeps that person's
+role and memberships as they were, rather than reading it as no groups, and
+logs an error saying so. "Groups assigned to the application" avoids this.
+So does SCIM below, which Entra also speaks: in the enterprise application
+under Provisioning, set the tenant URL to `https://<gateway>/scim/v2` and
+the secret token to the contents of `auth.scim.token_file`.
+
+## Active Directory
+
+With an on-premises Active Directory, people sign in on the gateway's own
+form with their AD username (`ada`, `CORP\ada`, `ada@corp.example`) and
+password. The gateway checks them with the domain controllers over LDAPS.
+
+The console sets this up and tries it first: `sudo llmgw`, People →
+Sign-in, **D**. It keeps the service account's password in the store
+(`bind_password_file: store:auth/ldap-bind`) and writes the CA into the
+settings (`ca:`), so nothing has to be copied to each control plane.
+By hand it is this block in the cluster's settings (`llmgw ctl config
+edit`), where files it names must then be on every control plane:
+
+```yaml
+auth:
+  enabled: true
+  ldap:
+    urls: [ldaps://dc1.corp.example, ldaps://dc2.corp.example]   # tried in order
+    ca_file: /etc/llmgw/corp-ca.pem         # the domain's CA, when not in the system's
+    bind_dn: CN=svc-llmgw,OU=Service,DC=corp,DC=example
+    bind_password_file: /etc/llmgw/secrets/ldap-bind   # root-only
+    base_dn: OU=Staff,DC=corp,DC=example    # where people are
+    user_filter: (memberOf:1.2.840.113556.1.4.1941:=CN=llmgw-users,OU=Groups,DC=corp,DC=example)
+    label: Acme                             # "Sign in with your Acme account"
+    create_accounts: true
+    roles:                                  # by group CN or full DN; nested groups count
+      admin: [llmgw-admins]
+      operator: [llmgw-ops]
+    memberships:
+      ml-engineers: "team:Acme/ml"
+    sync_interval: 15m                      # negative turns the sync off
+```
+
+- **Service account:** an ordinary domain user is enough: it only reads.
+  It finds the person and reads their groups. The person's password is
+  checked by binding as them, and is never stored.
+- **Encryption:** `ldaps://`, or `ldap://` with `start_tls: true`.
+  Plaintext LDAP is refused, since passwords would cross the network.
+- **Certificate:** the domain controller's certificate must name it in
+  its subject alternative names: a name in the CN only is refused, as are
+  certificates with a negative serial number. Samba's own autogenerated
+  certificate is both, so a Samba DC needs one from a CA (`tls certfile`
+  in smb.conf). A CA file a control plane cannot read turns directory
+  sign-in off on that control plane, said in its log. Local accounts still
+  sign in.
+- **Who may sign in:** enabled user objects under `base_dn` that match
+  `user_filter`. The example admits members of `llmgw-users`, nested groups
+  included.
+- **Accounts:** an account is tied to the person's `objectGUID`, so a changed
+  mail or name is still the same account. An account that already exists
+  with the person's mail is taken over. Without `create_accounts`, only
+  people who already have an account can sign in. Local accounts, such as
+  the administrator `server init` made, keep signing in with their own
+  password.
+- **Sync:** every `sync_interval` the leader reads the directory.
+  - Someone disabled or deleted in AD is suspended and their sessions
+    end.
+  - When they are enabled again they are reinstated. An account an
+    administrator suspended by hand stays suspended.
+  - Roles follow groups, and memberships the directory gave are taken away
+    when the person leaves the group.
+  - A sync that would suspend more than half of the directory's accounts is
+    refused and logged, rather than locking everyone out over a wrong
+    `base_dn` or filter.
+- **Passwords:** a person's password is changed in AD. The gateway sends no
+  reset link for these accounts.
+
 ## Secrets in Vault
 
 Every setting that names a file holding a secret — the store's `@address`,
@@ -2501,7 +3206,7 @@ loaded at boot by `llmgw-firewall.service`):
 | Machine | Ports, for the cluster only |
 | --- | --- |
 | control plane | 7443 (agents), metrics when not on loopback, 8081 (its mirror), engines and 9100 when it is a node, 2379/2380 when it holds the store's third etcd vote |
-| node | engine ports (`control.port_range`, 20000–20999), 9100 |
+| node | engine ports (`control.port_range`, 20000–20999), 9100; with a part in Slurm also 6817–6819 (slurmctld, slurmd, slurmdbd) and 60001–63000 (srun's `SrunPortRange`); exporting Slurm's shared files, 2049 (NFS) |
 | store member | 5432, 8008, 2379, 2380, 9100 |
 | mirror | 8081, 8082, 9100 |
 
@@ -3025,6 +3730,8 @@ requests. On a slow model, set `timeouts.generation` higher.
 
 - `docs/choosing-models.md` — which models to run and how many, if the answer
   to "we have GPUs, now what" is not already obvious
+- `docs/large-models.md` — a model on several cards or several nodes: how it
+  is split, what each way needs of a card, and bringing one up on a fabric
 - `docs/data.md` — what is stored and what leaves the machine, for the review
   that happens before this is allowed near anyone's documents
 - `docs/reference.md` — every configuration key, generated from the source, so

docs/knowledge.md

@@ -3,9 +3,9 @@
 A knowledge base is a collection of a company's documents that the cluster
 can answer from. People put their documents into it from their own machines
 with `llmgw-kb`. The tool reads the files on that machine, cuts them into
-passages ("chunks"), has the gateway embed each passage with the person's own
-key, and sends the text and the vectors. The files themselves never leave
-the machine.
+passages ("chunks") and sends their text; the gateway embeds each passage
+with the person's own key and keeps the text and the vectors. The files
+themselves never leave the machine.
 
 ## Why retrieval, and not training
 
@@ -159,10 +159,12 @@ llmgw ctl config set router.admission.bands '[{name: gold, priority: 300, satura
 llmgw keys create import --project Acme/docs --band bulk
 ```
 
-The people's questions then go first, and the import waits behind them. A
-request that waits as long as its band allows is answered 504 with
-`Retry-After`, and `llmgw-kb` sends it again rather than counting the
-passage as failed. It gives up after 15 minutes.
+The people's questions then go first, and the import waits behind them. The
+gateway embeds a document's passages 32 at a time, four batches at once,
+enough to fill an embedding replica's slots; a batch told to wait (`Retry-After`)
+waits up to five minutes before the document is answered the same, and
+`llmgw-kb` sends it again rather than counting it as failed. It gives up
+after 15 minutes.
 
 ## For everyone: `llmgw-kb`
 
@@ -225,6 +227,7 @@ not changed.
 | `--include`, `--exclude` | Glob patterns. `*.pdf` matches a file name; `reports/**/*.xlsx` matches a path. |
 | `--hidden` | Reads hidden files and folders too (left out by default). |
 | `-v` | Lists unchanged files too. |
+| `--embed-here` | Embeds the passages from this machine and sends the vectors. With a gateway that does not embed them itself it does so anyway. |
 
 Documents are named by their path under the folder, with forward slashes,
 on every system. With several folders, each folder's own name comes first.
@@ -259,13 +262,12 @@ scanned page with no OCR.
 
 ### What leaves your machine
 
-For each new or changed document, two things leave the machine, both over
-the gateway's TLS connection:
-
-- **The passages, to be embedded.** They go through `/v1/embeddings` like
-  any request of yours. Your key's limits, budget and guardrails apply.
-- **The document itself.** Its name, its file's SHA-256 digest and size,
-  and the passages with their vectors.
+For each new or changed document, one request leaves the machine, over the
+gateway's TLS connection: its name, its file's SHA-256 digest and size, and
+its passages' text. The gateway embeds the passages through `/v1/embeddings`
+as you, so your key's limits, budget and guardrails apply as if you had
+asked. That is about a kilobyte a passage; embedding here (`--embed-here`)
+brings each vector down and sends it up again, some thirty times more.
 
 The file itself is never sent. `--dry-run` shows exactly what would be.
 

docs/kong.md

@@ -0,0 +1,95 @@
+# llmgw and Kong
+
+For a site that already runs Kong. The two meet over plain HTTP: Kong in
+front of llmgw, as one more API it governs, or Kong behind llmgw, as an
+OpenAI-compatible provider. Nothing of either is built into the other.
+
+Tried with Kong Gateway 3.9 (`kong:3.9`, the last open-source release; 3.10
+and later need a Kong licence) in DB-less mode, in front of llmgw.
+
+## Kong in front of llmgw
+
+Two ways, both tried.
+
+**A plain route.** Kong forwards the request as it came. Tried with chat
+completions, streamed and not; llmgw's other APIs — Anthropic's
+`/v1/messages`, `/v1/responses`, embeddings, pictures — go the same way,
+since Kong does not read them. Kong adds its own authentication, rate limits
+and logging; llmgw still checks the key, the budget and the guardrails.
+
+```yaml
+_format_version: "3.0"
+services:
+- name: llmgw
+  url: https://llm.example.com          # llmgw's client port
+  routes:
+  - name: llmgw
+    paths: [/llmgw]                      # /llmgw/v1/chat/completions → /v1/chat/completions
+```
+
+**Kong's `ai-proxy` plugin**, for Kong's own token statistics on a chat
+route. llmgw is an OpenAI provider whose `upstream_url` is llmgw:
+
+```yaml
+services:
+- name: llmgw-chat
+  url: https://llm.example.com
+  routes:
+  - name: chat
+    paths: [/chat]
+  plugins:
+  - name: ai-proxy
+    config:
+      route_type: llm/v1/chat
+      logging: {log_statistics: true, log_payloads: false}
+      model:
+        provider: openai
+        name: qwen-coder                 # a model llmgw serves
+        options:
+          upstream_url: https://llm.example.com/v1/chat/completions
+```
+
+**Keys.** Without an `auth` block, `ai-proxy` passes the client's
+`Authorization` header on: each team keeps its own llmgw key, and llmgw's
+budgets, limits, guardrails and usage stay per team. With
+`auth: {header_name: Authorization, header_value: "Bearer sk-llmgw-…"}` Kong
+sends one key of its own instead, and llmgw sees every caller as that key —
+then budgets per team are Kong's to keep.
+
+**What Kong counts.** For a whole answer Kong reads llmgw's `usage` and
+records prompt and completion tokens as llmgw did. For a streamed answer
+Kong 3.9 records an estimate of the prompt tokens and no completion tokens,
+even with `stream_options.include_usage` set — the usage chunk reaches it
+and is not read. llmgw's own accounting counts both; use its usage export
+(`llmgw ctl usage export`) or `/admin/usage` for figures to bill by.
+
+**Timeouts.** A long answer from a model still loading outlasts Kong's
+60-second defaults: set the service's `read_timeout` to the time a model may
+take to start, as for nginx in [install.md](install.md).
+
+## Kong behind llmgw
+
+Not tried here. Kong's AI routes speak OpenAI's API, so llmgw reaches models Kong fronts as
+it reaches any provider: a model whose source is `external`.
+
+```yaml
+- name: kong-gpt
+  source: external
+  external:
+    url: https://kong.example.com/openai     # Kong's route, without /v1
+    model: gpt-4o-mini
+    api_key_file: /etc/llmgw/secrets/kong-consumer.key
+```
+
+The key is the one Kong's own authentication asks of a consumer. As a
+fallback (`fallbacks: [kong-gpt]`) it is used only when the cluster has no
+room.
+
+## Which to choose
+
+Kong in front when the site's API policy says every API goes through Kong:
+llmgw is then one more upstream, and its per-team accounting still holds
+when each team keeps its own key. Kong behind when the site already reaches
+outside providers through Kong and llmgw should too. llmgw does not need
+Kong for anything it does itself — keys, budgets, guardrails, routing to
+its own replicas.

docs/large-models.md

@@ -0,0 +1,145 @@
+# Large models: several cards, several nodes
+
+A model too big for one card runs on several. llmgw chooses how from the
+model's own figures and the cluster as it is; this page is the rules it
+follows, the commands, and what to look at when a first one is brought up
+on new hardware.
+
+## How a model is split
+
+In order, the first that fits:
+
+1. **One card.** Nothing to split.
+2. **Several cards of one node, tensor parallel.** Every layer is cut
+   across the cards, so they talk for every token: they must be on one
+   NVLink domain. The agent reads `nvidia-smi topo -m` and gives a replica
+   the cards with the most NVLink between them
+   ([install.md](install.md#nvlink-if-your-gpus-have-it)).
+3. **Several nodes, pipeline parallel between them.** Each node holds a
+   run of layers on all its cards (tensor parallel inside it), and only a
+   layer's output crosses to the next node. The nodes must share one
+   InfiniBand or RoCE fabric: a replica never spans two, and never a node
+   with none. Any number of nodes, not only two or four.
+
+The plan takes the fewest cards, then the fewest nodes, then the better
+precision up to eight bits.
+
+**Tensor width and the KV heads.** The cards split the model's KV heads
+among them, or past their number each card holds a copy of one. So the
+width divides the KV heads or is a multiple of them: a model with 8 KV
+heads runs 2, 4, 8 or 16 wide, not 12. Qwen3-235B has 4 and runs on 8.
+
+**Mixture of experts.** A model of experts (Qwen3-235B-A22B, DeepSeek-V3,
+Llama 4, Mixtral) on more than one card is served with vLLM's
+`--enable-expert-parallel`: each card holds whole experts rather than a
+slice of every one, which is less traffic per token. It is read from
+`config.json` (`num_experts`, `n_routed_experts`, `num_local_experts`).
+
+**When it does not fit.** A smaller precision first: an FP8 or AWQ
+repository of the same model, or a GGUF quantisation on llama.cpp if one
+node's cards hold it. Then more nodes. `llmgw ctl model plan REPO
+--precision bf16,fp8` shows each precision's fit from the model's config.
+
+## What a way needs
+
+Each way the plan offers says what a replica needs of each card, part by
+part, and of all its cards together:
+
+```
+$ llmgw ctl model plan meta-llama/Llama-3.3-70B-Instruct --format safetensors --context 32768
+FILE      SIZE    BEST WAY        A CARD   OF     ALL CARDS  NOW
+original  131 GB  4 × h100-80g    37.4 GB  72 GB  150 GB     free
+
+Best: original on 4 × h100-80g: 37.4 GB a card (weights 32.9 + context 2.5 + engine 2.0) of 72 usable; 150 GB on 4 cards.
+The context costs 0.31 GB per 1000 tokens, the whole model (--context to plan another).
+```
+
+- **weights:** the files, divided over the cards.
+- **context:** the KV cache for the context planned, divided likewise.
+- **engine:** the engine's own: 2 GB and a fiftieth of the weights for vLLM.
+- **usable:** 90% of a card; the rest is the driver's and fragmentation's.
+
+The console's **Serve a model** shows the same for every way a file runs.
+A context longer than planned is checked against it before anything is
+fetched.
+
+## Serving one
+
+Without flags the plan's best way is taken:
+
+```
+llmgw ctl model serve Qwen/Qwen3-235B-A22B-Instruct-2507-FP8 --format safetensors --context-length 32768
+```
+
+Or a way given by hand, for instance two nodes of eight:
+
+```
+llmgw ctl model serve meta-llama/Llama-3.3-70B-Instruct --format safetensors \
+    --gpu-class h100-80g --gpus 8 --nodes 2 --env NCCL_DEBUG=INFO
+```
+
+`--gpus` is the cards on each node and `--nodes` the nodes. The pipeline
+is the nodes and the tensor width the cards of each.
+
+`--env KEY=VALUE` goes into the engine's environment on every node:
+`NCCL_DEBUG=INFO` while bringing a cluster up, `NCCL_IB_*` to steer it.
+Only the variables engines are tuned with are taken. llmgw already sets
+these:
+- `NCCL_IB_HCA` to the fabric's adapters;
+- `NCCL_IB_GID_INDEX` on RoCE;
+- `NCCL_SOCKET_IFNAME` to the management network;
+- for an NVLink domain spanning nodes, `NCCL_MNNVL_ENABLE`.
+
+The job's first step says what it needs of each card. The weights then
+come to the mirror once and from it to each node. The startup timeout (20
+minutes unless the profile's `startup_timeout` says otherwise) runs from
+when a node's download last moved. So a model of hundreds of gigabytes is
+not failed while it is still arriving, and one whose download stops is
+failed that long after.
+
+## Watching it come up
+
+| What | Where |
+|---|---|
+| The replica and its shards, one per node, each with its download | `llmgw ctl replicas` |
+| Why it is not placed | `llmgw ctl explain-placement MODEL` |
+| A node's engine output (NCCL's lines with `NCCL_DEBUG=INFO`) | `llmgw ctl logs NODE --source SHARD` |
+| The bandwidth between two nodes, rail by rail | `llmgw ctl fabric-test NODE NODE` |
+| Each node's GPUs, NVLink and adapters | `llmgw doctor` on the node; its page under /admin |
+| Every step of a start, across nodes | the `replica start` trace ([install.md](install.md)) |
+
+With `NCCL_DEBUG=INFO`, NCCL says which transport it took between nodes:
+`NET/IB` is the fabric. `NET/Socket` means it fell back to TCP, and the
+model will run, slowly.
+
+## Bringing it up on new hardware
+
+A first run on a fabric, in order. Each step is worth its output if
+something goes wrong.
+
+1. **The nodes.** Add each with its fabric:
+   `llmgw node add host --fabric ib-a:mlx5_0,mlx5_1`. Then
+   `llmgw ctl nodes` shows every node ready, on `ib-a`.
+2. **The fabric.** `llmgw ctl fabric-test node1 node2`: each rail at its
+   rated speed (a 400 Gb/s rail near 45 GB/s).
+3. **One node, several cards.** A dense 70B at bf16 on four or eight
+   cards of one node, by the plan:
+   `llmgw ctl model serve meta-llama/Llama-3.3-70B-Instruct --format safetensors`.
+   Ready, then a question answered.
+4. **Two nodes.** The same model forced over two:
+   `--gpu-class CLASS --gpus 8 --nodes 2 --env NCCL_DEBUG=INFO`, under
+   another `--name`. Ready, answering, and `NET/IB` in both shards' logs.
+5. **Experts.** A mixture-of-experts model on one node's eight cards,
+   for instance `Qwen/Qwen3-235B-A22B-Instruct-2507-FP8`: its profile
+   carries `--enable-expert-parallel` (`llmgw ctl models`, or its
+   definition from the admin API).
+6. **A node lost.** With the two-node replica serving, stop one node's
+   agent (`systemctl stop llmgw-agent`): the replica fails and is placed
+   again once the node is back or another has room. Nothing is left
+   holding cards.
+
+For each, worth keeping:
+- the time from serve to ready;
+- tokens a second from `llmgw ctl bench`;
+- the plan's per-card figure against what `nvidia-smi` shows once loaded;
+- anything it said that was wrong.

docs/onboarding.md

@@ -54,10 +54,20 @@ on the page that creates it and never again.
 
 ### Connecting an editor
 
-The account page's **Connect your editor** makes a key for Continue, Codex or
-Cline and shows what the tool needs with that key, this gateway's address and a
-model already in it:
+The account page's **Connect your editor** makes a key for VS Code, Continue,
+Codex or Cline and shows what the tool needs with that key, this gateway's
+address and a model already in it:
 
+- **VS Code** — llmgw's own extension, downloaded from the gateway at
+  `/setup/vscode/llmgw.vsix` with this gateway's address already in it. It
+  offers every chat model the key may use to VS Code itself: the chat's model
+  picker, agent mode, and any extension that asks VS Code for a language model
+  (Cline's "VS Code LM API" provider among them), with no setup of their own.
+  **llmgw: Sign In** in VS Code opens this page with VS Code chosen; the page's
+  **Open in VS Code** link hands the key back, and the extension keeps it in
+  VS Code's secret storage. A link the extension did not ask for is confirmed
+  first, with the gateway's address shown, since it decides where code is sent.
+  **llmgw: Enter API Key** takes a pasted one instead.
 - **Continue** — an assistant file of its own, `~/.continue/assistants/llmgw.yaml`,
   so their `config.yaml` is not touched.
 - **Codex** — a provider and a profile in `~/.codex/config.toml`, run with
@@ -182,22 +192,41 @@ auth:
 A personal key has no project, so it keeps no conversations or agent notes;
 those are kept per project.
 
-### Organization managers
+### Organization managers and admins
 
-A member of an organization can be made its manager:
+A member of an organization can be made its manager, who sees it, or its
+admin, who also changes it:
 
 ```
 llmgw users join boss@acme.com --org Acme --role manager
+llmgw users join lead@acme.com --org Acme --role admin
 ```
 
-or with the star on their organization membership on `/admin/users`. A manager
-gets an **Organization** link and a page at `/org`: usage over time by team,
+or with the role chosen on their organization membership on `/admin/users`.
+Both get an **Organization** link and a page at `/org`: usage over time by team,
 project or person, the teams and projects with their limits and spend, and every
 member. Only what was spent in that organization's projects is on it; what its
-people do elsewhere is not. The page is read-only — a manager is not an
-administrator of the cluster, and limits and memberships stay with those who
-are. An administrator following `/org` lands on the same organization's panel
-on the people page instead, where all of that can be changed.
+people do elsewhere is not.
+
+A manager's page is read-only. An organization's admin also changes the
+organization from it:
+
+- adds and renames its teams and projects, and removes an empty one;
+- adds a person by their sign-in address to the organization, a team or a
+  project, and takes them out (their keys there are revoked with it);
+- makes a member of the organization a manager or an admin, or neither;
+- revokes a key spent in its projects.
+
+What reaches past the organization stays with the cluster's administrators:
+its limits and budgets, approving or suspending an account (someone not yet
+approved cannot be added), the organization's own name, and anything else on
+the cluster. An admin cannot take away their own role. A group from single
+sign-on or Active Directory that puts someone in the organization does not, by
+saying so again at the next sign-in, take a manager's or admin's role away.
+
+An administrator of the cluster following `/org` lands on the same
+organization's panel on the people page instead, where all of that can be
+changed.
 
 `llmgw keys init` still works for provisioning scripts and is idempotent, but it
 will no longer move a team or project that already exists under another parent;
@@ -507,8 +536,10 @@ included: `llmgw ctl audit`, or **Audit** in the console.
   suspends accounts from it; see [install.md](install.md#single-sign-on).
   Without them accounts are local, with local passwords.
 - **Staff roles are coarse.** `viewer`, `operator`, `billing` and `admin`
-  (`llmgw users role`) decide what someone may see and do under /admin; an
-  organization's manager can view it but not change it. Every change is in
-  the audit log (`llmgw ctl audit`).
+  (`llmgw users role`) decide what someone may see and do under /admin, and a
+  page draws only the forms that person may send: a viewer sees no forms. An
+  organization's manager can view it; its admin also changes its teams,
+  projects, people and keys. Every change is in the audit log (`llmgw ctl
+  audit`).
 - **Rate limits are per gateway instance.** With two instances each counts its own,
   so the effective limit doubles. v1 is single-instance by design.

docs/quick-install.md

@@ -8,7 +8,7 @@ explains every choice.
 
 | | |
 |---|---|
-| **A control plane** | One machine, or three for high availability (the store replicated, either one able to take over). Linux, x86-64 or ARM64: Debian 12, Ubuntu 22.04+ or RHEL 10. No GPU needed; 2 GB RAM, 20 GB disk each — 8 GB RAM when it is also a node running the starter model and the two judges. |
+| **A control plane** | One machine, or three for high availability (the store replicated, either one able to take over). Linux, x86-64 or ARM64: Debian 12, Ubuntu 22.04+ or RHEL 10. No GPU needed; 2 GB RAM, 20 GB disk each — 12 GB RAM when it is also a node running the starter model and the two judges, which setup checks before it installs anything; four cores or more, or the judges on machines of their own, keep a judgment within its wait. |
 | **GPU nodes** (optional to start) | The same Linux, with the NVIDIA driver installed. vLLM needs driver 525 or newer and a Turing GPU or newer; a V100 gets an older vLLM. |
 | **root** | on the control plane. On the other two of a three-machine setup, `sudo` (its password is asked for if it needs one); on nodes, `sudo` without a password. |
 | **SSH** | from the control plane to the other machines. The setup screens take a key, the ssh agent or a password; `node add` uses ssh keys. A machine nothing may log into joins with a code instead (`llmgw join`). |
@@ -146,7 +146,35 @@ agent is stopped.
 ## 4. Add a model
 
 Open `/admin/mirror` and choose **Explore models**. It lists what is popular on
-Hugging Face, showing only what your nodes can run.
+Hugging Face, showing only what your nodes can run: judged by the cluster as it
+is now — the largest card and its compute capability, the most memory on one
+node — so a model too big for them, in a form those cards lack (FP8 needs 8.9),
+or in neither GGUF nor safetensors is left out, and the page says how many.
+Without GPUs it lists GGUF only. Untick **only what these nodes can run** to
+see everything. On a model's page a file too big for the cluster is shown
+without a **Fetch** button, with what it would need.
+
+The **CPU tok/s** column is about how fast one reply would come on the CPU of
+the fastest node whose memory holds the model at Q4_K_M; a model's page gives
+it for each GGUF file by its own size. Every node's agent measures how fast its
+memory reads when it starts (it is in the agent's log as `memory_gbs`), and a
+token on a CPU reads every weight once: the estimate is that speed over the
+model's size — for a model of experts named like `30B-A3B`, its active part.
+On the test cluster it came within 15% of what llama.cpp measured. Several
+replies at once share the reads and each is slower.
+
+**GPU tok/s** is the same on the cards, for the layout this cluster would
+give the model — the fewest cards that hold it, of each class, the fastest of
+those; its tooltip says which ("2× h100-80g, NVLink, vLLM", "2 nodes × 1×
+a100-80g over ib0 (InfiniBand 200 Gb/s), vLLM"). A card's memory bandwidth is
+known from its model name (an H100 SXM and an H100 PCIe differ by half); a card
+not in the table gets no figure. Tensor parallel across a node's cards divides
+what each reads, and adds two all-reduces a layer, short over NVLink and
+longer over PCIe; llama.cpp across cards splits by layer, so the cards take
+turns and a reply is no faster than on one; a pipeline across nodes takes turns
+too, and adds a hop over the fabric for each node. One card's figures were
+measured on the test cluster's RTX 6000; tensor parallel and the fabric are
+from what NCCL is known to take, not measured here.
 
 1. Click a model to read its card.
 2. Pick one of its files, usually a `q4_k_m` for a first try, and click

docs/reference.md

@@ -37,6 +37,7 @@ rather than repeated.
 | `default_tpm` | `int` | [Auth](#auth) |
 | `signup_open` | `bool` | [Auth](#auth) |
 | `oidc` | `OIDC` | [Auth](#auth) |
+| `ldap` | `LDAP` | [Auth](#auth) |
 | `scim` | `SCIM` | [Auth](#auth) |
 | `max_key_ttl` | `time.Duration` | [Auth](#auth) |
 | `personal` | `Personal` | [Auth](#auth) |
@@ -52,6 +53,7 @@ rather than repeated.
 | `retain_days` | `int` | [Batch](#batch) |
 | `max_bytes` | `int64` | [Cache](#cache) |
 | `max_entry_bytes` | `int64` | [Cache](#cache) |
+| `embedding_model` | `string` | [Cache](#cache) |
 | `model` | `string` | [Canary](#canary) |
 | `percent` | `int` | [Canary](#canary) |
 | `min_requests` | `int` | [Canary](#canary) |
@@ -105,10 +107,13 @@ rather than repeated.
 | `url` | `string` | [External](#external) |
 | `model` | `string` | [External](#external) |
 | `api_key_file` | `string` | [External](#external) |
+| `api_key_files` | `[]ExternalKey` | [External](#external) |
 | `peer` | `string` | [External](#external) |
 | `api` | `string` | [External](#external) |
 | `workflow` | `string` | [External](#external) |
 | `output` | `string` | [External](#external) |
+| `file` | `string` | [ExternalKey](#externalkey) |
+| `weight` | `float64` | [ExternalKey](#externalkey) |
 | `budgets` | `[]FacilityBudget` | [Facility](#facility) |
 | `margin` | `float64` | [Facility](#facility) |
 | `gpu_watts` | `float64` | [Facility](#facility) |
@@ -121,6 +126,7 @@ rather than repeated.
 | `name` | `string` | [Federation](#federation) |
 | `interval` | `time.Duration` | [Federation](#federation) |
 | `enabled` | `bool` | [FirewallSettings](#firewallsettings) |
+| `binary` | `string` | [FreeTokenConfig](#freetokenconfig) |
 | `min_compute_capability` | `float64` | [GPUNeed](#gpuneed) |
 | `min_memory_gb` | `int` | [GPUNeed](#gpuneed) |
 | `format` | `string` | [GuardModel](#guardmodel) |
@@ -144,6 +150,21 @@ rather than repeated.
 | `tokenize_timeout` | `time.Duration` | [KVIndex](#kvindex) |
 | `min_blocks` | `int` | [KVIndex](#kvindex) |
 | `max_blocks` | `int` | [KVIndex](#kvindex) |
+| `urls` | `[]string` | [LDAP](#ldap) |
+| `start_tls` | `bool` | [LDAP](#ldap) |
+| `ca_file` | `string` | [LDAP](#ldap) |
+| `ca` | `string` | [LDAP](#ldap) |
+| `bind_dn` | `string` | [LDAP](#ldap) |
+| `bind_password_file` | `string` | [LDAP](#ldap) |
+| `base_dn` | `string` | [LDAP](#ldap) |
+| `group_base_dn` | `string` | [LDAP](#ldap) |
+| `user_filter` | `string` | [LDAP](#ldap) |
+| `label` | `string` | [LDAP](#ldap) |
+| `create_accounts` | `bool` | [LDAP](#ldap) |
+| `roles` | `map[string][]string` | [LDAP](#ldap) |
+| `memberships` | `map[string]string` | [LDAP](#ldap) |
+| `sync_interval` | `time.Duration` | [LDAP](#ldap) |
+| `timeout` | `time.Duration` | [LDAP](#ldap) |
 | `contact` | `string` | [Landing](#landing) |
 | `notice` | `string` | [Landing](#landing) |
 | `language` | `string` | [Landing](#landing) |
@@ -160,6 +181,7 @@ rather than repeated.
 | `parallel` | `int` | [LlamaCppConfig](#llamacppconfig) |
 | `threads` | `int` | [LlamaCppConfig](#llamacppconfig) |
 | `api` | `string` | [Mirror](#mirror) |
+| `peers` | `[]string` | [Mirror](#mirror) |
 | `tls` | `tlsconf.Config` | [Mirror](#mirror) |
 | `registry` | `string` | [Mirror](#mirror) |
 | `timeout` | `time.Duration` | [Mirror](#mirror) |
@@ -168,8 +190,10 @@ rather than repeated.
 | `aliases` | `[]string` | [Model](#model) |
 | `description` | `string` | [Model](#model) |
 | `source` | `string` | [Model](#model) |
+| `allow_pickle` | `bool` | [Model](#model) |
 | `profiles` | `[]Profile` | [Model](#model) |
 | `scaling` | `Scaling` | [Model](#model) |
+| `stopped` | `*Scaling` | [Model](#model) |
 | `class` | `string` | [Model](#model) |
 | `owner` | `string` | [Model](#model) |
 | `external` | `*External` | [Model](#model) |
@@ -211,6 +235,7 @@ rather than repeated.
 | `ib_root` | `string` | [Options](#options) |
 | `prefer_nvml` | `bool` | [Options](#options) |
 | `skip_gpus` | `bool` | [Options](#options) |
+| `dcgm_exporter` | `string` | [Options](#options) |
 | `repo` | `string` | [Origin](#origin) |
 | `revision` | `string` | [Origin](#origin) |
 | `files` | `[]OriginFile` | [Origin](#origin) |
@@ -238,6 +263,7 @@ rather than repeated.
 | `require_capabilities` | `bool` | [Policy](#policy) |
 | `validate_structured` | `bool` | [Policy](#policy) |
 | `judge` | `*JudgePolicy` | [Policy](#policy) |
+| `services` | `[]Service` | [Policy](#policy) |
 | `stream` | `string` | [Policy](#policy) |
 | `enabled` | `bool` | [Prefix](#prefix) |
 | `max_nodes` | `int` | [Prefix](#prefix) |
@@ -269,6 +295,7 @@ rather than repeated.
 | `keep_within` | `string` | [Profile](#profile) |
 | `weight` | `float64` | [Profile](#profile) |
 | `priority` | `int` | [Profile](#profile) |
+| `replicas` | `int` | [Profile](#profile) |
 | `startup_timeout` | `time.Duration` | [Profile](#profile) |
 | `kv_events` | `bool` | [Profile](#profile) |
 | `gpu` | `*GPUNeed` | [Profile](#profile) |
@@ -294,6 +321,7 @@ rather than repeated.
 | `batch` | `Batch` | [Router](#router) |
 | `cache` | `Cache` | [Router](#router) |
 | `trusted_proxies` | `[]string` | [Router](#router) |
+| `active_active` | `bool` | [Router](#router) |
 | `timeouts` | `Timeouts` | [Router](#router) |
 | `scoring` | `Scoring` | [Router](#router) |
 | `admission` | `Admission` | [Router](#router) |
@@ -301,6 +329,7 @@ rather than repeated.
 | `injection` | `string` | [Rules](#rules) |
 | `deny` | `[]Pattern` | [Rules](#rules) |
 | `max_chars` | `int` | [Rules](#rules) |
+| `restore` | `bool` | [Rules](#rules) |
 | `token_file` | `string` | [SCIM](#scim) |
 | `python` | `string` | [SGLangConfig](#sglangconfig) |
 | `ttft_p99` | `time.Duration` | [SLO](#slo) |
@@ -347,6 +376,7 @@ rather than repeated.
 | `mail` | `mail.Config` | [Server](#server) |
 | `serving` | `ServingSettings` | [Server](#server) |
 | `firewall` | `FirewallSettings` | [Server](#server) |
+| `slurm` | `SlurmSettings` | [Server](#server) |
 | `presets` | `PresetSettings` | [Server](#server) |
 | `on_demand` | `OnDemandSettings` | [Server](#server) |
 | `agents` | `AgentSettings` | [Server](#server) |
@@ -361,8 +391,33 @@ rather than repeated.
 | `facility` | `Facility` | [Server](#server) |
 | `federation` | `Federation` | [Server](#server) |
 | `endpoints` | `[]Endpoint` | [Server](#server) |
+| `name` | `string` | [Service](#service) |
+| `url` | `string` | [Service](#service) |
+| `header_files` | `map[string]string` | [Service](#service) |
+| `body` | `string` | [Service](#service) |
+| `blocked` | `string` | [Service](#service) |
+| `reason` | `string` | [Service](#service) |
+| `on` | `string` | [Service](#service) |
+| `action` | `string` | [Service](#service) |
+| `timeout` | `time.Duration` | [Service](#service) |
 | `stopped` | `bool` | [ServingSettings](#servingsettings) |
+| `path` | `string` | [SharedStorage](#sharedstorage) |
+| `nfs` | `string` | [SharedStorage](#sharedstorage) |
+| `server` | `string` | [SharedStorage](#sharedstorage) |
 | `bin` | `string` | [SimConfig](#simconfig) |
+| `enabled` | `bool` | [SlurmSettings](#slurmsettings) |
+| `cluster_name` | `string` | [SlurmSettings](#slurmsettings) |
+| `controller` | `string` | [SlurmSettings](#slurmsettings) |
+| `nodes` | `[]string` | [SlurmSettings](#slurmsettings) |
+| `all_nodes` | `bool` | [SlurmSettings](#slurmsettings) |
+| `lent` | `[]string` | [SlurmSettings](#slurmsettings) |
+| `partition` | `string` | [SlurmSettings](#slurmsettings) |
+| `shared` | `SharedStorage` | [SlurmSettings](#slurmsettings) |
+| `users` | `string` | [SlurmSettings](#slurmsettings) |
+| `containers` | `bool` | [SlurmSettings](#slurmsettings) |
+| `dir` | `string` | [StrataConfig](#strataconfig) |
+| `python` | `string` | [StrataConfig](#strataconfig) |
+| `engine` | `string` | [StrataConfig](#strataconfig) |
 | `python` | `string` | [TRTLLMConfig](#trtllmconfig) |
 | `store` | `string` | [Threads](#threads) |
 | `retain_hours` | `int` | [Threads](#threads) |
@@ -394,6 +449,7 @@ rather than repeated.
 | `secret_file` | `string` | [Webhook](#webhook) |
 | `prompt_per_1k` | `int64` | [accounting.Price](#accountingprice) |
 | `completion_per_1k` | `int64` | [accounting.Price](#accountingprice) |
+| `cached_per_1k` | `int64` | [accounting.Price](#accountingprice) |
 | `per_image` | `int64` | [accounting.Price](#accountingprice) |
 | `per_mesh` | `int64` | [accounting.Price](#accountingprice) |
 | `name` | `string` | [admission.Band](#admissionband) |
@@ -432,6 +488,7 @@ rather than repeated.
 | `queue_ttl` | `time.Duration` | [config.Band](#configband) |
 | `prompt_per_1k` | `int64` | [config.Price](#configprice) |
 | `completion_per_1k` | `int64` | [config.Price](#configprice) |
+| `cached_per_1k` | `int64` | [config.Price](#configprice) |
 | `per_image` | `int64` | [config.Price](#configprice) |
 | `per_mesh` | `int64` | [config.Price](#configprice) |
 | `disabled` | `[]string` | [engines.Config](#enginesconfig) |
@@ -441,6 +498,8 @@ rather than repeated.
 | `trtllm` | `TRTLLMConfig` | [engines.Config](#enginesconfig) |
 | `sim` | `SimConfig` | [engines.Config](#enginesconfig) |
 | `comfyui` | `ComfyUIConfig` | [engines.Config](#enginesconfig) |
+| `strata` | `StrataConfig` | [engines.Config](#enginesconfig) |
+| `freetoken` | `FreeTokenConfig` | [engines.Config](#enginesconfig) |
 | `enabled` | `*bool` | [guard.Config](#guardconfig) |
 | `default` | `[]string` | [guard.Config](#guardconfig) |
 | `policies` | `map[string]Policy` | [guard.Config](#guardconfig) |
@@ -616,6 +675,15 @@ OIDC is sign-in through an identity provider.
 
 Takes a [OIDC](#oidc) block.
 
+### `ldap`
+
+`LDAP`
+
+LDAP is sign-in against a directory, Active Directory above all, and
+the accounts kept in step with it.
+
+Takes a [LDAP](#ldap) block.
+
 ### `scim`
 
 `SCIM`
@@ -728,6 +796,15 @@ Past it, the answers used longest ago go first.
 MaxEntryBytes bounds one answer; a longer one is not kept. 1 MB
 unless set.
 
+### `embedding_model`
+
+`string`
+
+EmbeddingModel embeds the questions of projects that reuse answers to
+similar ones (llmgw keys cache --similar): an embedding model in the
+catalog. Unset, those projects reuse only the same request's answer.
+The gateway asks it on its own account, as it asks a judge.
+
 ## Canary
 
 ### `model`
@@ -948,7 +1025,7 @@ work stands without being renewed.
 Only meaningful with a shared store, where a second control plane can
 see the claim. Long enough to survive a slow query or a brief network
 stall, short enough that a standby takes over before anyone has finished
-reading the alert. Zero takes the default.
+reading the alert. Zero takes the default, 10s.
 
 ## CustomNode
 
@@ -1103,6 +1180,15 @@ APIKeyFile holds the provider's key, on the control plane. A file
 rather than the key itself, so the key is in neither the catalog, the
 store nor the audit log; empty for a provider that asks for none.
 
+### `api_key_files`
+
+`[]ExternalKey`
+
+APIKeyFiles are several keys for the provider, in place of
+APIKeyFile: one endpoint each (external:MODEL#1, #2…), so a key that
+answers 429 or 401 is passed over while its breaker is open and the
+request goes to another. A provider's limits are per key.
+
 ### `peer`
 
 `string`
@@ -1135,6 +1221,20 @@ plane.
 
 Output is what it makes: pictures, or meshes (see ComfyUI.Output).
 
+## ExternalKey
+
+### `file`
+
+`string`
+
+File holds the key, on the control plane, as APIKeyFile does.
+
+### `weight`
+
+`float64`
+
+Weight is its share of the requests against the others'; zero is 1.
+
 ## Facility
 
 ### `budgets`
@@ -1235,6 +1335,16 @@ site whose own firewall does that job; the rules are kept, and true
 puts them back. Changed in the cluster's settings, it is in force on
 every node and control plane within seconds.
 
+## FreeTokenConfig
+
+### `binary`
+
+`string`
+
+Binary is its ft command, in the environment it was installed into.
+The agent installs one itself on a node whose GPUs it runs on; this
+is for one installed by hand.
+
 ## GPUNeed
 
 ### `min_compute_capability`
@@ -1401,6 +1511,114 @@ MaxBlocks bounds how many of a prompt's blocks are hashed and looked
 up. Beyond it a longer match still helps, but the difference no longer
 decides anything. Zero takes 4096.
 
+## LDAP
+
+### `urls`
+
+`[]string`
+
+URLs are the directory's servers, tried in turn: ldaps:// (636), or
+ldap:// (389) with StartTLS.
+
+### `start_tls`
+
+`bool`
+
+StartTLS upgrades an ldap:// connection before anything is sent. A
+password never crosses a connection without TLS: ldap:// without it
+is refused.
+
+### `ca_file`
+
+`string`
+
+CAFile is the authority the servers' certificates are checked against,
+when it is not the system's: an AD Certificate Services root.
+
+### `ca`
+
+`string`
+
+CA is the same authority written out, PEM: kept in the cluster's
+settings, so no file has to be put on every control plane. Used in
+place of CAFile when both are set.
+
+### `bind_dn`
+
+`string`
+
+BindDN and the file holding its password: a service account that may
+read users and groups. Nothing is written to the directory.
+
+### `bind_password_file`
+
+`string`
+
+### `base_dn`
+
+`string`
+
+BaseDN is where people are searched for, subtree included.
+
+### `group_base_dn`
+
+`string`
+
+GroupBaseDN is where their groups are; empty is BaseDN's domain root.
+
+### `user_filter`
+
+`string`
+
+UserFilter narrows who may sign in, ANDed with the login match: a
+group (memberOf=...) or an OU's people only. Empty is every enabled
+user object.
+
+### `label`
+
+`string`
+
+Label is what the form says to sign in with: "Sign in with your
+<label> account". Empty is "Active Directory".
+
+### `create_accounts`
+
+`bool`
+
+CreateAccounts makes an account, active, for a person the directory
+vouches for and the gateway does not know. Off, only existing accounts
+sign in this way.
+
+### `roles`
+
+`map[string][]string`
+
+Roles maps a staff role (viewer, operator, billing, admin) to the
+groups that give it, by name (CN) or whole DN. Set at every sign-in
+and sync: someone taken out of the group loses the role.
+
+### `memberships`
+
+`map[string]string`
+
+Memberships maps a group, by CN or DN, to a place in the tree —
+org:NAME, team:ORG/TEAM, project:ORG/TEAM/PROJECT — joined at sign-in
+and kept by the sync, which also takes away a place whose group the
+person has left.
+
+### `sync_interval`
+
+`time.Duration`
+
+SyncInterval is how often the leader reads the directory. Zero is
+15m; negative turns the sync off, leaving only sign-in.
+
+### `timeout`
+
+`time.Duration`
+
+Timeout bounds each round trip to a server. Zero is 10s.
+
 ## Landing
 
 ### `contact`
@@ -1539,6 +1757,14 @@ The management port, not the one nodes fetch bytes from. Those are two
 listeners on the mirror precisely so this one can be locked down without
 re-certifying every agent's fetcher.
 
+### `peers`
+
+`[]string`
+
+Peers are the management addresses of mirrors kept in step with this
+one (llmgw mirror serve --peer): asked when it cannot be reached, and
+a node given the same file on one of them while it is down.
+
 ### `tls`
 
 `tlsconf.Config`
@@ -1608,6 +1834,16 @@ choice by guesswork.
 
 Source is the weights path or a HuggingFace id.
 
+### `allow_pickle`
+
+`bool`
+
+AllowPickle lets the nodes fetch this model's weights, or its
+adapters', in a pickle format (.bin, .pt, .pth, .ckpt), which run code
+as they are loaded. Off, such a source is refused when the model is
+added and by the node before it fetches anything. For a model that has
+no safetensors or GGUF form and whose files were checked.
+
 ### `profiles`
 
 `[]Profile`
@@ -1618,6 +1854,16 @@ Source is the weights path or a HuggingFace id.
 
 Takes a [Scaling](#scaling) block.
 
+### `stopped`
+
+`*Scaling`
+
+Stopped is the scaling a model had when an administrator stopped it:
+kept in the catalog with none running (Scaling 0/0) until started
+again with this. Nil for a model not stopped.
+
+Takes a [Scaling](#scaling) block.
+
 ### `class`
 
 `string`
@@ -1965,6 +2211,15 @@ is used, is the more exact source of the peers.
 SkipGPUs omits GPU discovery. A node without nvidia-smi is already
 found to have none, so this is for one whose driver is broken.
 
+### `dcgm_exporter`
+
+`string`
+
+DCGMExporter is the dcgm-exporter metrics URL read for how much of
+each card's chip is in use, on cards NVML's GPM does not cover (before
+Hopper). Empty is http://127.0.0.1:9400/metrics; "off" reads none. A
+node without one costs a refused connection every five minutes.
+
 ## Origin
 
 ### `repo`
@@ -2136,6 +2391,12 @@ Judge has a judge model on the cluster read the text too.
 
 Takes a [JudgePolicy](#judgepolicy) block.
 
+### `services`
+
+`[]Service`
+
+Services are outside guardrails asked over HTTP, beside the judge.
+
 ### `stream`
 
 `string`
@@ -2365,6 +2626,16 @@ rather than a guess (plan §6.2.1).
 
 Priority orders profiles during placement; highest that fits wins.
 
+### `replicas`
+
+`int`
+
+Replicas, set on any of a model's profiles, is how many of its
+replicas run on this one: seven on cards and four on CPUs, all
+answering under the model's name, rather than each replica on the
+best profile that fits. The model's scaling is then the sum, min and
+max alike; a profile left at zero runs none.
+
 ### `startup_timeout`
 
 `time.Duration`
@@ -2565,6 +2836,22 @@ as addresses or CIDR prefixes. Their X-Forwarded-For, -Proto and -Host
 are believed; anyone else's are not, since they decide the address
 written into the setup scripts and whose attempts are counted where.
 
+### `active_active`
+
+`bool`
+
+ActiveActive has standby control planes answer the client API
+themselves, routing to the leader's replicas, rather than passing
+every request to the leader. More machines carry the traffic, which
+is the point. Rate limits are the cluster's: each router's counts
+reach the others through the leader every two seconds, so a key
+spread over N control planes may pass its limit by what they allow
+in those two seconds. A budget is checked against the store every
+half minute; cached answers and prefix affinity are split between
+the routers. Off by
+default. Which machine a client reaches is the load balancer's or
+DNS's to say; the floating address stays on the leader.
+
 ### `timeouts`
 
 `Timeouts`
@@ -2610,6 +2897,14 @@ Deny are patterns, matched against the text as a model reads it.
 
 MaxChars refuses text longer than this; zero is no limit.
 
+### `restore`
+
+`bool`
+
+Restore, on input, puts what redact rewrote as placeholders back into
+the answer: the model reads "[EMAIL_1]", the client reads its own
+address. Only the values that request sent, only into its answer.
+
 ## SCIM
 
 ### `token_file`
@@ -2998,6 +3293,15 @@ Firewall turns llmgw's firewall on every machine on or off.
 
 Takes a [FirewallSettings](#firewallsettings) block.
 
+### `slurm`
+
+`SlurmSettings`
+
+Slurm is a Slurm cluster run beside llmgw on some of its nodes, and
+which of them it may use now.
+
+Takes a [SlurmSettings](#slurmsettings) block.
+
 ### `presets`
 
 `PresetSettings`
@@ -3117,6 +3421,67 @@ Endpoints is the phase-1 static replica list. It is only consulted when
 no model declares a profile, which is what makes a router-only smoke test
 against sim-engines possible without a control plane.
 
+## Service
+
+### `name`
+
+`string`
+
+Name says which service decided, in the verdict's rule.
+
+### `url`
+
+`string`
+
+URL is POSTed the body.
+
+### `header_files`
+
+`map[string]string`
+
+HeaderFiles are headers whose values are read from files each time,
+so a rotated key needs no restart: {Authorization: /etc/…/lakera}.
+
+### `body`
+
+`string`
+
+Body is the request, a template given .Text; {{json .Text}} writes it
+as a JSON string. Empty is {"input": <text>}.
+
+### `blocked`
+
+`string`
+
+Blocked is where in the answer the service says the text is unsafe:
+a dotted path, numbers for list items (results.0.flagged). True, a
+number above zero, a non-empty list or string other than "false" is
+unsafe.
+
+### `reason`
+
+`string`
+
+Reason is where it says why, for the verdict; optional.
+
+### `on`
+
+`string`
+
+On: input, output or both. Empty is input.
+
+### `action`
+
+`string`
+
+Action on unsafe: block or flag. Empty is block.
+
+### `timeout`
+
+`time.Duration`
+
+Timeout bounds one call. Empty is 3s; past it the policy's fail says.
+
 ## ServingSettings
 
 ### `stopped`
@@ -3129,6 +3494,30 @@ running and every model's definition as it is: llmgw stop sets it,
 llmgw start clears it, and the models come back as their catalog
 entries ask. In force on every control plane within seconds.
 
+## SharedStorage
+
+### `path`
+
+`string`
+
+Path is where every machine sees it, as /shared.
+
+### `nfs`
+
+`string`
+
+NFS is the site's export, host:/path, mounted at Path on every
+machine of the Slurm cluster.
+
+### `server`
+
+`string`
+
+Server is the llmgw node, the controller or one of the nodes, that
+exports Path over NFS to the others: for a site with no storage of
+its own. One machine: not highly available, and no faster than its
+disk and network.
+
 ## SimConfig
 
 ### `bin`
@@ -3138,6 +3527,109 @@ entries ask. In force on every control plane within seconds.
 Bin is the llmgw binary itself; the engine is one of its subcommands.
 Empty means whatever started this agent, which is right almost always.
 
+## SlurmSettings
+
+### `enabled`
+
+`bool`
+
+Enabled runs Slurm on the nodes below. False stops llmgw keeping it;
+what was installed is left installed, and every node is llmgw's.
+
+### `cluster_name`
+
+`string`
+
+ClusterName is Slurm's ClusterName. Empty is "llmgw".
+
+### `controller`
+
+`string`
+
+Controller is the llmgw node whose agent runs slurmctld. It needs no
+GPU, and need not be one of Nodes.
+
+### `nodes`
+
+`[]string`
+
+Nodes are the llmgw nodes that run slurmd. Each stays llmgw's, drained
+in Slurm, until it is also in Lent.
+
+### `all_nodes`
+
+`bool`
+
+AllNodes runs slurmd on every llmgw node, those that join later too,
+besides any named in Nodes. Each is llmgw's until it is lent, as any.
+
+### `lent`
+
+`[]string`
+
+Lent are the nodes of Nodes Slurm may run jobs on now. A node put here
+is drained of llmgw's replicas first, as `ctl drain` does, and resumed
+in Slurm once none is left; a node taken out is drained in Slurm and
+given back to llmgw when its last job ends.
+
+### `partition`
+
+`string`
+
+Partition is the one partition every node is in. Empty is "gpu".
+
+### `shared`
+
+`SharedStorage`
+
+Shared is the file system every machine of the Slurm cluster sees at
+the same path: home directories, datasets, the jobs' output and
+checkpoints. Empty path: none, and a batch job's output stays on the
+node it ran on.
+
+Takes a [SharedStorage](#sharedstorage) block.
+
+### `users`
+
+`string`
+
+Users is who keeps the Linux users that run jobs, the same on every
+machine of the Slurm cluster: "llmgw" makes one for each active llmgw
+account (its login from the e-mail address, a UID of llmgw's from
+200000 up, its SSH keys from the portal), let in to the controller,
+and to a node only while a job of theirs runs there (pam_slurm_adopt);
+empty leaves users to the site (a directory through sssd, or by hand).
+
+### `containers`
+
+`bool`
+
+Containers lets jobs run in containers: `srun --container-image
+nvcr.io#nvidia/pytorch:25.01-py3` (pyxis, with enroot under it),
+the GPUs and the shared path inside. The agents install both on
+every Slurm machine; jobs take it once every machine has them.
+
+## StrataConfig
+
+### `dir`
+
+`string`
+
+Dir is Strata's source directory, where setup.py and serve/ are.
+
+### `python`
+
+`string`
+
+Python is the interpreter of its environment; Dir's .venv when unset.
+
+### `engine`
+
+`string`
+
+Engine is the directory of its prebuilt engine, where the strata
+binary and the CUDA libraries it ships are; Dir's engine/ when unset.
+
 ## TRTLLMConfig
 
 ### `python`
@@ -3355,6 +3847,15 @@ is sha256=<hex HMAC of the body>.
 
 `int64`
 
+### `cached_per_1k`
+
+`int64`
+
+CachedPer1k is what a prompt token costs that the engine or provider
+read from its cache instead of computing: the part of prompt_tokens
+the answer counts as cached_tokens. Zero charges those at
+PromptPer1k, as before there was a rate for them.
+
 ### `per_image`
 
 `int64`
@@ -3629,6 +4130,14 @@ had rather wait behind people than be turned away.
 
 `int64`
 
+### `cached_per_1k`
+
+`int64`
+
+CachedPer1k is the cost of a prompt token the engine or provider read
+from its cache (the answer's cached_tokens, counted within the
+prompt's). Zero charges them at prompt_per_1k.
+
 ### `per_image`
 
 `int64`
@@ -3693,6 +4202,25 @@ ComfyUI makes pictures; offered by a node whose ComfyUI is named here.
 
 Takes a [ComfyUIConfig](#comfyuiconfig) block.
 
+### `strata`
+
+`StrataConfig`
+
+Strata serves a few mixture-of-experts models from RAM and one card;
+offered by a node whose Strata is named here, or that installed it.
+
+Takes a [StrataConfig](#strataconfig) block.
+
+### `freetoken`
+
+`FreeTokenConfig`
+
+FreeToken serves mixture-of-experts models larger than a card from
+RAM and one card of 8.0 or newer; installed by the agent where the
+cards allow it, or named here.
+
+Takes a [FreeTokenConfig](#freetokenconfig) block.
+
 ## guard.Config
 
 ### `enabled`

go.mod

@@ -12,6 +12,7 @@ require (
 	github.com/charmbracelet/x/ansi v0.11.6
 	github.com/coreos/go-oidc/v3 v3.21.0
 	github.com/go-jose/go-jose/v4 v4.1.4
+	github.com/go-ldap/ldap/v3 v3.4.15
 	github.com/go-zeromq/zmq4 v0.17.0
 	github.com/jackc/pgx/v5 v5.11.0
 	github.com/knadh/koanf/parsers/yaml v1.1.1
@@ -35,6 +36,7 @@ require (
 	golang.org/x/net v0.59.0
 	golang.org/x/oauth2 v0.37.0
 	golang.org/x/sync v0.23.0
+	golang.org/x/sys v0.48.0
 	golang.org/x/term v0.46.0
 	golang.org/x/text v0.42.0
 	google.golang.org/protobuf v1.36.12
@@ -44,6 +46,7 @@ require (
 require (
 	filippo.io/edwards25519 v1.2.0 // indirect
 	filippo.io/hpke v0.4.0 // indirect
+	github.com/Azure/go-ntlmssp v0.1.1 // indirect
 	github.com/aymanbagabas/go-osc52/v2 v2.0.1 // indirect
 	github.com/beorn7/perks v1.0.1 // indirect
 	github.com/cenkalti/backoff/v5 v5.0.3 // indirect
@@ -56,6 +59,7 @@ require (
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f // indirect
 	github.com/fsnotify/fsnotify v1.9.0 // indirect
+	github.com/go-asn1-ber/asn1-ber v1.5.8 // indirect
 	github.com/go-logr/logr v1.4.4 // indirect
 	github.com/go-logr/stdr v1.2.2 // indirect
 	github.com/go-viper/mapstructure/v2 v2.5.0 // indirect
@@ -88,7 +92,6 @@ require (
 	go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.46.0 // indirect
 	go.opentelemetry.io/otel/metric v1.46.0 // indirect
 	go.opentelemetry.io/proto/otlp v1.11.0 // indirect
-	golang.org/x/sys v0.48.0 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20260819154853-08b0e4226688 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20260819154853-08b0e4226688 // indirect
 	google.golang.org/grpc v1.83.1 // indirect

go.sum

@@ -8,8 +8,12 @@ filippo.io/edwards25519 v1.2.0 h1:crnVqOiS4jqYleHd9vaKZ+HKtHfllngJIiOpNpoJsjo=
 filippo.io/edwards25519 v1.2.0/go.mod h1:xzAOLCNug/yB62zG1bQ8uziwrIqIuxhctzJT18Q77mc=
 filippo.io/hpke v0.4.0 h1:p575VVQ6ted4pL+it6M00V/f2qTZITO0zgmdKCkd5+A=
 filippo.io/hpke v0.4.0/go.mod h1:EmAN849/P3qdeK+PCMkDpDm83vRHM5cDipBJ8xbQLVY=
+github.com/Azure/go-ntlmssp v0.1.1 h1:l+FM/EEMb0U9QZE7mKNEDw5Mu3mFiaa2GKOoTSsNDPw=
+github.com/Azure/go-ntlmssp v0.1.1/go.mod h1:NYqdhxd/8aAct/s4qSYZEerdPuH1liG2/X9DiVTbhpk=
 github.com/NVIDIA/go-nvml v0.13.4-0 h1:o3jp9u2x1R9ShFE3v+Aesp55XOSIQFMJz/VGNUcJaNE=
 github.com/NVIDIA/go-nvml v0.13.4-0/go.mod h1:id63qwpoDWpFXwnwM6psDCSqW4BmNu6mWpr4YeQtPGo=
+github.com/alexbrainman/sspi v0.0.0-20250919150558-7d374ff0d59e h1:4dAU9FXIyQktpoUAgOJK3OTFc/xug0PCXYCqU0FgDKI=
+github.com/alexbrainman/sspi v0.0.0-20250919150558-7d374ff0d59e/go.mod h1:cEWa1LVoE5KvSD9ONXsZrj0z6KqySlCCNKHlLzbqAt4=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1 h1:HwpRHbFMcZLEVr42D4p7XBqjyuxQH5SMiErDT4WkJ2k=
 github.com/aymanbagabas/go-osc52/v2 v2.0.1/go.mod h1:uYgXzlJ7ZpABp8OJ+exZzJJhRNQ2ASbcXHWsFqH8hp8=
 github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
@@ -46,8 +50,12 @@ github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f h1:Y/CXytFA4m6
 github.com/erikgeiser/coninput v0.0.0-20211004153227-1c3628e74d0f/go.mod h1:vw97MGsxSvLiUE2X8qFplwetxpGLQrlU1Q9AUEIzCaM=
 github.com/fsnotify/fsnotify v1.9.0 h1:2Ml+OJNzbYCTzsxtv8vKSFD9PbJjmhYF14k/jKC7S9k=
 github.com/fsnotify/fsnotify v1.9.0/go.mod h1:8jBTzvmWwFyi3Pb8djgCCO5IBqzKJ/Jwo8TRcHyHii0=
+github.com/go-asn1-ber/asn1-ber v1.5.8 h1:H9AZkK22UOmfX8J84ubyaZxKJZ3FMHVwn8swoMML7iQ=
+github.com/go-asn1-ber/asn1-ber v1.5.8/go.mod h1:hEBeB/ic+5LoWskz+yKT7vGhhPYkProFKoKdwZRWMe0=
 github.com/go-jose/go-jose/v4 v4.1.4 h1:moDMcTHmvE6Groj34emNPLs/qtYXRVcd6S7NHbHz3kA=
 github.com/go-jose/go-jose/v4 v4.1.4/go.mod h1:x4oUasVrzR7071A4TnHLGSPpNOm2a21K9Kf04k1rs08=
+github.com/go-ldap/ldap/v3 v3.4.15 h1:eRB4f4Ecn1POL1mO86+uUDlecmyWsdiJwh60bAUS4UY=
+github.com/go-ldap/ldap/v3 v3.4.15/go.mod h1:jeD0c98EdZQCKKKKe4kTWs5aCf6y+j8KuYsYLznZSQk=
 github.com/go-logr/logr v1.2.2/go.mod h1:jdQByPbusPIv2/zmleS9BjJVeZ6kBagPoEUsqbVz/1A=
 github.com/go-logr/logr v1.4.4 h1:tG4xh9yMsRCAiodLVTxyrkzSZ9+o0L1Kg/+cPVcbP/8=
 github.com/go-logr/logr v1.4.4/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ40KvY=
@@ -69,6 +77,8 @@ github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
 github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
 github.com/grpc-ecosystem/grpc-gateway/v2 v2.30.0 h1:/Tnpcb2E0Pz/tN9s3bfEY2Q8ePCEX9iuS+cneUwncnw=
 github.com/grpc-ecosystem/grpc-gateway/v2 v2.30.0/go.mod h1:zOBXOsUaBSjKgmH4OGzV1esUpR3oUSCPYVd2cUBjKYY=
+github.com/hashicorp/go-uuid v1.0.3 h1:2gKiV6YVmrJ1i2CKKa9obLvRieoRGviZFL26PcT/Co8=
+github.com/hashicorp/go-uuid v1.0.3/go.mod h1:6SBZvOh/SIDV7/2o3Jml5SYk/TvGqwFJ/bN7x4byOro=
 github.com/hashicorp/golang-lru/v2 v2.0.7 h1:a+bsQ5rvGLjzHuww6tVxozPZFVghXaHOwFs4luLUK2k=
 github.com/hashicorp/golang-lru/v2 v2.0.7/go.mod h1:QeFd9opnmA6QUJc5vARoKUSoFhyfM2/ZepoAG6RGpeM=
 github.com/inconshreveable/mousetrap v1.1.0 h1:wN+x4NVGpMsO7ErUn/mUI3vEoE6Jt13X2s0bqwp9tc8=
@@ -81,6 +91,18 @@ github.com/jackc/pgx/v5 v5.11.0 h1:IzBBtyK9AHqf98cctWFifYSci2hgQR/cd56wB4p+ogg=
 github.com/jackc/pgx/v5 v5.11.0/go.mod h1:mal1tBGAFfLHvZzaYh77YS/eC6IX9OWbRV1QIIM0Jn4=
 github.com/jackc/puddle/v2 v2.2.2 h1:PR8nw+E/1w0GLuRFSmiioY6UooMp6KJv0/61nB7icHo=
 github.com/jackc/puddle/v2 v2.2.2/go.mod h1:vriiEXHvEE654aYKXXjOvZM39qJ0q+azkZFrfEOc3H4=
+github.com/jcmturner/aescts/v2 v2.0.0 h1:9YKLH6ey7H4eDBXW8khjYslgyqG2xZikXP0EQFKrle8=
+github.com/jcmturner/aescts/v2 v2.0.0/go.mod h1:AiaICIRyfYg35RUkr8yESTqvSy7csK90qZ5xfvvsoNs=
+github.com/jcmturner/dnsutils/v2 v2.0.0 h1:lltnkeZGL0wILNvrNiVCR6Ro5PGU/SeBvVO/8c/iPbo=
+github.com/jcmturner/dnsutils/v2 v2.0.0/go.mod h1:b0TnjGOvI/n42bZa+hmXL+kFJZsFT7G4t3HTlQ184QM=
+github.com/jcmturner/gofork v1.7.6 h1:QH0l3hzAU1tfT3rZCnW5zXl+orbkNMMRGJfdJjHVETg=
+github.com/jcmturner/gofork v1.7.6/go.mod h1:1622LH6i/EZqLloHfE7IeZ0uEJwMSUyQ/nDd82IeqRo=
+github.com/jcmturner/goidentity/v6 v6.0.1 h1:VKnZd2oEIMorCTsFBnJWbExfNN7yZr3EhJAxwOkZg6o=
+github.com/jcmturner/goidentity/v6 v6.0.1/go.mod h1:X1YW3bgtvwAXju7V3LCIMpY0Gbxyjn/mY9zx4tFonSg=
+github.com/jcmturner/gokrb5/v8 v8.4.4 h1:x1Sv4HaTpepFkXbt2IkL29DXRf8sOfZXo8eRKh687T8=
+github.com/jcmturner/gokrb5/v8 v8.4.4/go.mod h1:1btQEpgT6k+unzCwX1KdWMEwPPkkgBtP+F6aCACiMrs=
+github.com/jcmturner/rpc/v2 v2.0.3 h1:7FXXj8Ti1IaVFpSAziCZWNzbNuZmnvw/i6CqLNdWfZY=
+github.com/jcmturner/rpc/v2 v2.0.3/go.mod h1:VUJYCIDm3PVOEHw8sgt091/20OJjskO/YJki3ELg/Hc=
 github.com/klauspost/compress v1.19.1 h1:VsB4HPswih7mmZ8WleSFQ75c/Ui1M4trX5oAsJnhSlk=
 github.com/klauspost/compress v1.19.1/go.mod h1:cwPg85FWrGar70rWktvGQj8/hthj3wpl0PGDogxkrSQ=
 github.com/knadh/koanf/maps v0.1.2 h1:RBfmAW5CnZT+PJ1CVc1QSJKf4Xu9kxfQgYVQSu8hpbo=

internal/account/directory_test.go

@@ -0,0 +1,63 @@
+package account
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/store"
+)
+
+type fakeDirectory struct {
+	asked []string
+	u     store.User
+	err   error
+}
+
+func (f *fakeDirectory) SignIn(_ context.Context, login, _ string) (store.User, error) {
+	f.asked = append(f.asked, login)
+	return f.u, f.err
+}
+
+// With a directory, a name the store has no password for is the
+// directory's to check; an account with a password here — the
+// administrator init made — is still checked here, so a directory that is
+// down locks nobody out.
+func TestPasswordsKeptElsewhereAreTheDirectorysToCheck(t *testing.T) {
+	svc, st := newService(t)
+	ctx := t.Context()
+	hash, _ := HashPassword("local-secret")
+	admin := store.User{ID: "usr-admin", Email: "admin@example.com", Name: "a", PasswordHash: hash,
+		State: store.UserActive, CreatedAt: time.Now()}
+	ada := store.User{ID: "usr-ada", Email: "ada@corp.example", Name: "Ada", PasswordHash: "!ldap",
+		State: store.UserActive, CreatedAt: time.Now()}
+	for _, u := range []store.User{admin, ada} {
+		if err := st.CreateUser(ctx, u); err != nil {
+			t.Fatal(err)
+		}
+	}
+	dir := &fakeDirectory{u: ada}
+	svc.SetDirectory(dir)
+
+	if _, _, err := svc.SignIn(ctx, "admin@example.com", "local-secret"); err != nil || len(dir.asked) != 0 {
+		t.Fatalf("the local administrator: %v, directory asked %v", err, dir.asked)
+	}
+	// A name, not an address, and an account kept by the directory.
+	for _, login := range []string{`CORP\ada`, "ada@corp.example"} {
+		if tok, _, err := svc.SignIn(ctx, login, "ad-secret"); err != nil || tok == "" {
+			t.Errorf("%s: %v", login, err)
+		}
+	}
+	if len(dir.asked) != 2 || dir.asked[0] != `CORP\ada` {
+		t.Errorf("the directory was asked %v", dir.asked)
+	}
+	dir.err = ErrBadCredentials
+	if _, _, err := svc.SignIn(ctx, "ada", "wrong"); !errors.Is(err, ErrBadCredentials) {
+		t.Errorf("a password the directory refused: %v", err)
+	}
+	// A reset link would give a directory's account a password here.
+	if _, _, err := svc.RequestReset(ctx, st, "ada@corp.example"); !errors.Is(err, ErrNoReset) {
+		t.Errorf("a directory's account was sent a reset link: %v", err)
+	}
+}

internal/account/service.go

@@ -80,8 +80,26 @@ type Service struct {
 
 	mu       sync.Mutex
 	attempts map[string]*attemptWindow
+
+	// dir checks the passwords of accounts this store keeps none for.
+	dir Directory
 }
 
+// Directory checks a password against an outside directory — Active
+// Directory over LDAP — and answers the account to sign in: made or kept in
+// step with the directory. Its errors are this package's ErrBadCredentials
+// and ErrSuspended for the person's own, and anything else for the
+// directory's.
+type Directory interface {
+	SignIn(ctx context.Context, login, password string) (store.User, error)
+}
+
+// SetDirectory has passwords for accounts with none here, and names that are
+// not addresses, checked by the directory. An account with a password here
+// — the administrator init made — is still checked here, so a directory that
+// is down locks nobody out.
+func (s *Service) SetDirectory(d Directory) { s.dir = d }
+
 type attemptWindow struct {
 	count int
 	since time.Time
@@ -137,6 +155,7 @@ func (s *Service) SignUp(ctx context.Context, email, name, password string) erro
 // SignIn checks a password and opens a session, returning the session token and
 // the CSRF token that must accompany the forms in it.
 func (s *Service) SignIn(ctx context.Context, email, password string) (token, csrf string, err error) {
+	login := strings.TrimSpace(email)
 	email = store.NormalizeEmail(email)
 	// Counted per address and the client trying it, so somebody guessing
 	// from elsewhere cannot lock the owner out of their own account; and per
@@ -151,6 +170,15 @@ func (s *Service) SignIn(ctx context.Context, email, password string) (token, cs
 	}
 
 	u, err := s.store.UserByEmail(ctx, email)
+	// No password kept here — the directory's, or nobody's — and a
+	// directory to ask: it decides.
+	if s.dir != nil && (errors.Is(err, store.ErrNotFound) || (err == nil && strings.HasPrefix(u.PasswordHash, "!"))) {
+		if u, err = s.dir.SignIn(ctx, login, password); err != nil {
+			return "", "", err
+		}
+		s.clearAttempts(byClient)
+		return s.open(ctx, u)
+	}
 	if errors.Is(err, store.ErrNotFound) {
 		// Verify anyway, against a hash that cannot match. Returning early here
 		// would let a stopwatch enumerate every account on the system.
@@ -180,7 +208,11 @@ func (s *Service) SignIn(ctx context.Context, email, password string) (token, cs
 	case store.UserSuspended:
 		return "", "", ErrSuspended
 	}
+	return s.open(ctx, u)
+}
 
+// open makes a session for a person whose password was right.
+func (s *Service) open(ctx context.Context, u store.User) (token, csrf string, err error) {
 	if token, err = store.NewSessionToken(); err != nil {
 		return "", "", err
 	}
@@ -402,7 +434,10 @@ func (s *Service) RequestReset(ctx context.Context, rs ResetStore, email string)
 		return "", store.User{}, ErrTooManyTries
 	}
 	u, err := rs.UserByEmail(ctx, email)
-	if err != nil || u.PasswordHash == "" || u.State != store.UserActive {
+	// An account whose password is kept elsewhere — "!oidc", "!scim",
+	// "!ldap" — gets no link: setting one here would give it a second way
+	// in that its directory knows nothing of, and outlives its leaving.
+	if err != nil || u.PasswordHash == "" || strings.HasPrefix(u.PasswordHash, "!") || u.State != store.UserActive {
 		return "", store.User{}, ErrNoReset
 	}
 	token, err := rs.CreatePasswordReset(ctx, u.ID, ResetLinkTTL)

internal/accounting/accounting.go

@@ -78,10 +78,41 @@ func (a *Accountant) RestoreBudgets(ctx context.Context, s store.Store, period s
 	if start.IsZero() {
 		return nil
 	}
-	rows, err := s.SpendSinceByScope(ctx, start)
+	spend, keys, err := storedSpend(ctx, s, start)
 	if err != nil {
 		return fmt.Errorf("restoring budgets: %w", err)
 	}
+	a.budgets.Restore(period, spend)
+	a.log.Info("restored budgets from the store",
+		"keys", keys, "counters", len(spend), "since", start)
+	return nil
+}
+
+// CatchUpBudgets raises each counter to what the store has recorded, where
+// that is more. Run every so often by a router that is one of several taking
+// requests: the others' spend reaches it only through the store. Never
+// lowers a counter, which leads the store by whatever this process has not
+// written yet.
+func (a *Accountant) CatchUpBudgets(ctx context.Context, s store.Store, period store.Period) error {
+	start := PeriodStart(period, time.Now())
+	if start.IsZero() {
+		return nil
+	}
+	spend, _, err := storedSpend(ctx, s, start)
+	if err != nil {
+		return fmt.Errorf("reading budgets: %w", err)
+	}
+	a.budgets.Raise(period, spend)
+	return nil
+}
+
+// storedSpend is the spend since start the store has recorded, by counter,
+// and how many keys it came from.
+func storedSpend(ctx context.Context, s store.Store, start time.Time) (map[string]int64, int, error) {
+	rows, err := s.SpendSinceByScope(ctx, start)
+	if err != nil {
+		return nil, 0, err
+	}
 	// Every scope a key sits under, not just the key. A budget owned by a team
 	// is one counter fed by every key beneath it, and restoring only per-key
 	// totals would bring the control plane back up believing the team had spent
@@ -113,10 +144,7 @@ func (a *Accountant) RestoreBudgets(ctx context.Context, s store.Store, period s
 			}
 		}
 	}
-	a.budgets.Restore(period, spend)
-	a.log.Info("restored budgets from the store",
-		"keys", len(rows), "counters", len(spend), "since", start)
-	return nil
+	return spend, len(rows), nil
 }
 
 // Reservation is what Check returns: the estimate it charged, so Record can
@@ -323,7 +351,7 @@ func (a *Accountant) Record(res Reservation, u store.Usage) {
 	if u.ServedModel != "" && u.ServedModel != u.Model && a.pricing != nil {
 		price = a.pricing.For(u.ServedModel)
 	}
-	u.CostMicros = price.Cost(u.PromptTokens, u.CompletionTokens, u.Images, u.Meshes)
+	u.CostMicros = price.Cost(u.PromptTokens, u.CachedTokens, u.CompletionTokens, u.Images, u.Meshes)
 	for _, b := range res.Budgets {
 		a.budgets.Add(b.Counter, b.Period, u.CostMicros)
 	}

internal/accounting/accounting_test.go

@@ -39,20 +39,37 @@ func TestCostRoundsUpSoSmallRequestsAreNotFree(t *testing.T) {
 	// Rounding down would make a stream of tiny requests cost nothing, which is
 	// exactly the traffic shape a quota exists to bound.
 	p := Price{PromptPer1k: 1000, CompletionPer1k: 3000}
-	if got := p.Cost(1, 0, 0, 0); got != 1 {
+	if got := p.Cost(1, 0, 0, 0, 0); got != 1 {
 		t.Errorf("one prompt token cost %d", got)
 	}
-	if got := p.Cost(1000, 1000, 0, 0); got != 4000 {
+	if got := p.Cost(1000, 0, 1000, 0, 0); got != 4000 {
 		t.Errorf("a thousand of each cost %d, want 4000", got)
 	}
-	if got := p.Cost(0, 0, 0, 0); got != 0 {
+	if got := p.Cost(0, 0, 0, 0, 0); got != 0 {
 		t.Errorf("an empty request cost %d", got)
 	}
 }
 
+func TestACachedPromptCostsTheCachedRate(t *testing.T) {
+	p := Price{PromptPer1k: 1000, CachedPer1k: 100, CompletionPer1k: 3000}
+	// 10,000 prompt tokens of which 9,000 were read from the cache: 1,000 at
+	// the prompt rate and 9,000 at the cached one.
+	if got, want := p.Cost(10000, 9000, 0, 0, 0), int64(1000+900); got != want {
+		t.Errorf("a 90%% cached prompt cost %d, want %d", got, want)
+	}
+	// More cached than prompt is an engine's miscount, not a refund.
+	if got := p.Cost(100, 500, 0, 0, 0); got != 10 {
+		t.Errorf("cached beyond the prompt cost %d, want 10", got)
+	}
+	// No cached rate: cached tokens cost what any prompt token does.
+	if got := (Price{PromptPer1k: 1000}).Cost(10000, 9000, 0, 0, 0); got != 10000 {
+		t.Errorf("without a cached rate a cached prompt cost %d, want 10000", got)
+	}
+}
+
 func TestAPictureCostsItsPriceOnTopOfItsPrompt(t *testing.T) {
 	p := Price{PromptPer1k: 1000, PerImage: 50000}
-	if got := p.Cost(20, 0, 3, 0); got != 20+150000 {
+	if got := p.Cost(20, 0, 0, 3, 0); got != 20+150000 {
 		t.Errorf("three pictures from a 20-token prompt cost %d", got)
 	}
 	if (Price{PerImage: 1}).Zero() {
@@ -199,6 +216,22 @@ func TestBudgetSurvivesARestart(t *testing.T) {
 	}
 }
 
+func TestCatchingUpRaisesButNeverLowers(t *testing.T) {
+	// Another router's spend arrives through the store; this one's own, not
+	// written yet, must not be wiped by a store that is behind.
+	b := NewBudgets()
+	b.Add("mine", store.PeriodMonth, 900)
+	b.Add("shared", store.PeriodMonth, 100)
+
+	b.Raise(store.PeriodMonth, map[string]int64{"mine": 400, "shared": 700, "theirs": 50})
+
+	for key, want := range map[string]int64{"mine": 900, "shared": 700, "theirs": 50} {
+		if got := b.Spent(key, store.PeriodMonth); got != want {
+			t.Errorf("%s: spent %d, want %d", key, got, want)
+		}
+	}
+}
+
 func TestBudgetsResetAtThePeriodBoundary(t *testing.T) {
 	b := NewBudgets()
 	now := time.Date(2026, 1, 31, 23, 59, 0, 0, time.UTC)

internal/accounting/budget.go

@@ -61,6 +61,19 @@ func (b *Budgets) Restore(period store.Period, spend map[string]int64) {
 	}
 }
 
+// Raise lifts counters to the spend given, where it is more.
+func (b *Budgets) Raise(period store.Period, spend map[string]int64) {
+	start := PeriodStart(period, b.now())
+	b.mu.Lock()
+	defer b.mu.Unlock()
+	for key, micros := range spend {
+		if c := b.currentLocked(key, start); micros > c.micros {
+			c.micros = micros
+			b.spend[key] = c
+		}
+	}
+}
+
 // Spent returns a key's spend in the current period, resetting it if the period
 // has rolled over.
 func (b *Budgets) Spent(keyID string, period store.Period) int64 {

internal/accounting/cost.go

@@ -18,6 +18,11 @@ import (
 type Price struct {
 	PromptPer1k     int64 `koanf:"prompt_per_1k"`
 	CompletionPer1k int64 `koanf:"completion_per_1k"`
+	// CachedPer1k is what a prompt token costs that the engine or provider
+	// read from its cache instead of computing: the part of prompt_tokens
+	// the answer counts as cached_tokens. Zero charges those at
+	// PromptPer1k, as before there was a rate for them.
+	CachedPer1k int64 `koanf:"cached_per_1k"`
 	// PerImage is what one picture costs, for a model that makes them: a
 	// picture's cost is the card's time, which its prompt's few tokens do
 	// not measure.
@@ -33,15 +38,24 @@ type Price struct {
 // whose budget is spent, which is what makes the plan's budget fallback work
 // (§7.7).
 func (p Price) Zero() bool {
-	return p.PromptPer1k == 0 && p.CompletionPer1k == 0 && p.PerImage == 0 && p.PerMesh == 0
+	return p.PromptPer1k == 0 && p.CompletionPer1k == 0 && p.CachedPer1k == 0 && p.PerImage == 0 && p.PerMesh == 0
 }
 
 // Cost computes what a request spent, rounding up.
 //
 // Up rather than to nearest: rounding down would make a stream of tiny requests
 // free, which is exactly the shape of traffic a quota exists to bound.
-func (p Price) Cost(promptTokens, completionTokens, images, meshes int64) int64 {
-	return divUp(promptTokens*p.PromptPer1k, 1000) + divUp(completionTokens*p.CompletionPer1k, 1000) +
+//
+// cachedTokens are counted within promptTokens, as OpenAI's usage counts
+// them, and are charged at the cached rate instead of the prompt rate.
+func (p Price) Cost(promptTokens, cachedTokens, completionTokens, images, meshes int64) int64 {
+	cachedTokens = min(max(cachedTokens, 0), max(promptTokens, 0))
+	cachedRate := p.CachedPer1k
+	if cachedRate == 0 {
+		cachedRate = p.PromptPer1k
+	}
+	return divUp((promptTokens-cachedTokens)*p.PromptPer1k+cachedTokens*cachedRate, 1000) +
+		divUp(completionTokens*p.CompletionPer1k, 1000) +
 		max(images, 0)*p.PerImage + max(meshes, 0)*p.PerMesh
 }
 

internal/accounting/limiter.go

@@ -25,7 +25,11 @@ type Limiter struct {
 type window struct {
 	// current and previous are counts in this minute and the one before.
 	current, previous int64
-	start             time.Time
+	// othersCurrent and othersPrevious are the same counts on the other
+	// control planes taking requests, as last heard: what makes a limit the
+	// cluster's rather than each process's.
+	othersCurrent, othersPrevious int64
+	start                         time.Time
 }
 
 // NewLimiter builds an empty limiter.
@@ -104,21 +108,24 @@ func (w *window) roll(now time.Time) {
 		return
 	case start.Sub(w.start) == windowLen:
 		w.previous, w.current = w.current, 0
+		w.othersPrevious, w.othersCurrent = w.othersCurrent, 0
 	default:
 		// More than one window has passed with no traffic.
 		w.previous, w.current = 0, 0
+		w.othersPrevious, w.othersCurrent = 0, 0
 	}
 	w.start = start
 }
 
 // rate weights the previous window by how much of it is still in view.
 func (w *window) rate(now time.Time) float64 {
+	cur, prev := w.current+w.othersCurrent, w.previous+w.othersPrevious
 	elapsed := now.Sub(w.start)
 	if elapsed >= windowLen {
-		return float64(w.current)
+		return float64(cur)
 	}
 	remaining := 1 - float64(elapsed)/float64(windowLen)
-	return float64(w.previous)*remaining + float64(w.current)
+	return float64(prev)*remaining + float64(cur)
 }
 
 // Forget drops a key's window. Used when a key is revoked; keeping it would

internal/accounting/shared.go

@@ -0,0 +1,110 @@
+package accounting
+
+import "time"
+
+// Counts are one limiter's counts by counter: this minute's and the one
+// before, as absolute figures, so reporting them twice changes nothing.
+type Counts struct {
+	Minute   time.Time        `json:"minute"`
+	Current  map[string]int64 `json:"current,omitempty"`
+	Previous map[string]int64 `json:"previous,omitempty"`
+}
+
+// Add sums other into c, aligned to c's minute: counts of a minute that is
+// c's previous one land there, older ones are dropped. Clocks a second apart
+// at a minute's turn are why.
+func (c *Counts) Add(other Counts) {
+	switch {
+	case other.Minute.Equal(c.Minute):
+		c.Current = addInto(c.Current, other.Current)
+		c.Previous = addInto(c.Previous, other.Previous)
+	case other.Minute.Add(windowLen).Equal(c.Minute):
+		c.Previous = addInto(c.Previous, other.Current)
+	}
+}
+
+func addInto(dst, src map[string]int64) map[string]int64 {
+	if len(src) == 0 {
+		return dst
+	}
+	if dst == nil {
+		dst = make(map[string]int64, len(src))
+	}
+	for k, v := range src {
+		dst[k] += v
+	}
+	return dst
+}
+
+// Counts reads this limiter's own counts, without the others'.
+func (l *Limiter) Counts() Counts {
+	now := l.now()
+	out := Counts{Minute: now.Truncate(windowLen), Current: map[string]int64{}, Previous: map[string]int64{}}
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	for k, w := range l.keys {
+		w.roll(now)
+		if w.current > 0 {
+			out.Current[k] = w.current
+		}
+		if w.previous > 0 {
+			out.Previous[k] = w.previous
+		}
+	}
+	return out
+}
+
+// SetOthers replaces what the other control planes have counted. A counter
+// they did not mention has nothing from them.
+func (l *Limiter) SetOthers(c Counts) {
+	now := l.now()
+	aligned := Counts{Minute: now.Truncate(windowLen)}
+	aligned.Add(c)
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	for k, w := range l.keys {
+		w.roll(now)
+		w.othersCurrent, w.othersPrevious = aligned.Current[k], aligned.Previous[k]
+		delete(aligned.Current, k)
+		delete(aligned.Previous, k)
+	}
+	// Counters only others have used yet: a key spread over the cluster is
+	// limited here from its first request.
+	for k := range aligned.Current {
+		w := l.windowLocked(k, now)
+		w.othersCurrent, w.othersPrevious = aligned.Current[k], aligned.Previous[k]
+		delete(aligned.Previous, k)
+	}
+	for k, v := range aligned.Previous {
+		l.windowLocked(k, now).othersPrevious = v
+	}
+}
+
+// Rates are an accountant's two limiters' counts.
+type Rates struct {
+	Requests Counts `json:"requests"`
+	Tokens   Counts `json:"tokens"`
+}
+
+// Add sums other into r.
+func (r *Rates) Add(other Rates) {
+	r.Requests.Add(other.Requests)
+	r.Tokens.Add(other.Tokens)
+}
+
+// NewRates is an empty sum for the minute now is in.
+func NewRates(now time.Time) Rates {
+	m := now.Truncate(windowLen)
+	return Rates{Requests: Counts{Minute: m}, Tokens: Counts{Minute: m}}
+}
+
+// Rates reads this process's own request and token counts.
+func (a *Accountant) Rates() Rates {
+	return Rates{Requests: a.requests.Counts(), Tokens: a.tokens.Counts()}
+}
+
+// SetOtherRates puts the other control planes' counts in force here.
+func (a *Accountant) SetOtherRates(r Rates) {
+	a.requests.SetOthers(r.Requests)
+	a.tokens.SetOthers(r.Tokens)
+}

internal/accounting/shared_test.go

@@ -0,0 +1,71 @@
+package accounting
+
+import (
+	"testing"
+	"time"
+)
+
+// Two control planes each allowing a key six of its ten a minute: once each
+// has heard the other's count, neither allows another.
+func TestALimitIsTheClustersOnceTheCountsAreShared(t *testing.T) {
+	now := time.Date(2026, 11, 1, 12, 0, 10, 0, time.UTC)
+	a, b := NewLimiter(), NewLimiter()
+	a.now, b.now = func() time.Time { return now }, func() time.Time { return now }
+	for range 6 {
+		if !a.Allow("k", 10, 1) || !b.Allow("k", 10, 1) {
+			t.Fatal("refused within the limit")
+		}
+	}
+	a.SetOthers(b.Counts())
+	b.SetOthers(a.Counts())
+	if a.Allow("k", 10, 1) || b.Allow("k", 10, 1) {
+		t.Error("twelve of ten allowed once both knew")
+	}
+	// Reporting the same counts again changes nothing: they are absolute.
+	a.SetOthers(b.Counts())
+	if got := a.Rate("k"); got != 12 {
+		t.Errorf("rate %v after the same counts twice", got)
+	}
+	// What the others counted is never charged back to them.
+	if c := a.Counts(); c.Current["k"] != 6 {
+		t.Errorf("own count %d", c.Current["k"])
+	}
+}
+
+// A key only the others have used is limited here from its first request,
+// and their counts move into the previous minute with the clock.
+func TestOthersCountsFollowTheMinute(t *testing.T) {
+	now := time.Date(2026, 11, 1, 12, 0, 50, 0, time.UTC)
+	a, b := NewLimiter(), NewLimiter()
+	a.now, b.now = func() time.Time { return now }, func() time.Time { return now }
+	for range 10 {
+		b.Allow("k", 10, 1)
+	}
+	a.SetOthers(b.Counts())
+	if a.Allow("k", 10, 1) {
+		t.Error("a key the cluster has used up was allowed where it had not been seen")
+	}
+	now = now.Add(20 * time.Second) // 12:01:10, a sixth into the next minute
+	if got := a.Rate("k"); got < 8 || got > 8.5 {
+		t.Errorf("rate %v: the others' last minute should weigh five sixths", got)
+	}
+	// Counts of a minute ago, sent late, land in the previous minute.
+	old := b.Counts()
+	old.Minute = now.Truncate(time.Minute).Add(-time.Minute)
+	old.Current, old.Previous = map[string]int64{"k": 3}, nil
+	a.SetOthers(old)
+	if got := a.Rate("k"); got < 2 || got > 3 {
+		t.Errorf("rate %v from late counts", got)
+	}
+}
+
+func TestRatesAddAlignsMinutes(t *testing.T) {
+	m := time.Date(2026, 11, 1, 12, 1, 0, 0, time.UTC)
+	sum := NewRates(m)
+	sum.Add(Rates{Requests: Counts{Minute: m, Current: map[string]int64{"k": 2}, Previous: map[string]int64{"k": 1}}})
+	sum.Add(Rates{Requests: Counts{Minute: m.Add(-time.Minute), Current: map[string]int64{"k": 5}}})
+	sum.Add(Rates{Requests: Counts{Minute: m.Add(-2 * time.Minute), Current: map[string]int64{"k": 100}}})
+	if sum.Requests.Current["k"] != 2 || sum.Requests.Previous["k"] != 6 {
+		t.Errorf("sum %+v", sum.Requests)
+	}
+}

internal/agent/agent.go

@@ -12,6 +12,7 @@ import (
 	"io"
 	"log/slog"
 	"maps"
+	"math"
 	"math/rand/v2"
 	"net"
 	"net/http"
@@ -34,6 +35,7 @@ import (
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/engines"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/provision"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/slurmnode"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/supervisor"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/train"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/weights"
@@ -46,7 +48,9 @@ import (
 
 // Agent runs on a GPU node.
 type Agent struct {
-	cfg Config
+	// upgrading is true while an upgrade downloads.
+	upgrading atomic.Bool
+	cfg       Config
 	// answered says the current session heard from its control plane, which
 	// starts the reconnect backoff again when it ends.
 	answered atomic.Bool
@@ -94,7 +98,10 @@ type Agent struct {
 	// (nil: this machine's shell).
 	fwMu     sync.Mutex
 	fwRunner firewall.Runner
-	inv      discovery.Inventory
+	// fwLast is the last firewall message, applied again when the node
+	// is given a part in Slurm. Under fwMu.
+	fwLast *llmgwv1.Firewall
+	inv    discovery.Inventory
 	// shardReplica remembers which replica a shard belongs to, so a status
 	// report can name both. The control plane keys on the replica.
 	shardReplica map[string]string
@@ -145,6 +152,10 @@ type Agent struct {
 
 	// trainer runs LoRA training jobs, one at a time (BACKLOG F4).
 	trainer *train.Runner
+	// slurm keeps the node's part in Slurm (BACKLOG P); slurmOut is where
+	// its status goes, under mu.
+	slurm    *slurmnode.Keeper
+	slurmOut chan<- *llmgwv1.AgentMessage
 }
 
 // connectHTTPClient is the HTTP client connect uses; an interface so tests can
@@ -184,7 +195,7 @@ func New(cfg Config, httpClient connectHTTPClient, log *slog.Logger) (*Agent, er
 		logRing:      ring,
 		provKick:     make(chan struct{}, 1),
 		load:         discovery.NewLoadSampler(),
-		sensors:      discovery.NewSensorSampler(),
+		sensors:      discovery.NewSensorSampler().WithProfile(discovery.NewProfSampler(cfg.Discovery.DCGMExporter)),
 		wear:         discovery.NewHealthSampler(),
 		serverList:   loadControlPlanes(cfg.Servers, cfg.TLS.Enabled()),
 		kv:           kvevents.NewMirror(),
@@ -203,6 +214,7 @@ func New(cfg Config, httpClient connectHTTPClient, log *slog.Logger) (*Agent, er
 	}
 	a.trainer = train.New(train.Env{EnginesDir: enginesDir, WorkDir: filepath.Join(cfg.Superv.WorkDir, "trainer"),
 		RunAs: runAs, Log: log})
+	a.slurm = slurmnode.New(slurmnode.Local{}, a.prov, slurmMarker, log)
 	return a, nil
 }
 
@@ -229,6 +241,7 @@ func (a *Agent) Run(ctx context.Context) error {
 	}
 	go a.provisionLoop(ctx)
 	go a.ensureDiags(ctx)
+	go a.slurm.Run(ctx, a.sendSlurm)
 
 	attempt := 0
 	next := 0
@@ -392,8 +405,11 @@ func (a *Agent) settleAdopted(sh *supervisor.Shard, o supervisor.Orphan) {
 		_ = a.sup.Stop(ctx, o.ShardID)
 		return
 	}
-	// What it computes on, for one started before the agent kept that.
-	if a.sup.ComputeOf(o.ShardID) == nil {
+	// What it computes on, for one started before the agent kept that — or
+	// kept as the CPU for a GPU placement by an agent that did not follow
+	// an engine into a process group of its own, which the check above has
+	// just found on its GPUs.
+	if c := a.sup.ComputeOf(o.ShardID); c == nil || (len(o.Spec.GPUIndices) > 0 && c.Backend == "CPU") {
 		if c, ok := a.gpuCompute(ctx, o.Spec.GPUIndices, sh.Status().PID); ok {
 			a.sup.SetCompute(o.ShardID, c)
 		}
@@ -513,6 +529,8 @@ func (a *Agent) scan(ctx context.Context) error {
 	inv, err := discovery.Discover(ctx, opts, a.log)
 	host := discovery.DiscoverHost()
 	host.MinComputeCapability = discovery.MinComputeCapability(inv.GPUs)
+	bw := a.memBandwidth()
+	host.MemBandwidthGBs, host.MemBandwidthCoreGBs = bw.All, bw.Core
 	a.mu.Lock()
 	a.inv = inv
 	a.prov.Host = host
@@ -522,6 +540,18 @@ func (a *Agent) scan(ctx context.Context) error {
 	return err
 }
 
+// forgetSent is a new stream: a control plane that knows nothing of where
+// installs stand, nor which files are here beyond what registration says —
+// and registration says what the startup scan found. So the next heartbeat
+// tells it both again. Without it, files deleted after the agent started came
+// back on a control plane that restarted, and stayed until the next change.
+func (a *Agent) forgetSent() {
+	a.mu.Lock()
+	defer a.mu.Unlock()
+	a.provSent = ""
+	a.lastWeights = ""
+}
+
 // session runs one connection until it breaks.
 func (a *Agent) session(ctx context.Context, server string) error {
 	client := llmgwv1connect.NewAgentServiceClient(a.http, server)
@@ -558,17 +588,19 @@ func (a *Agent) session(ctx context.Context, server string) error {
 	reg.EngineVersions = a.engines(parent)
 	reg.EnginesProbed = true
 	reg.InstallableEngines = slices.DeleteFunc(provision.Installable(), a.disabled)
-	reg.Features = []string{FeatureFabricTest}
+	reg.Features = []string{FeatureFabricTest, FeatureSlurm}
 	reg.RunningShards = a.runningShards()
 	a.mu.Lock()
 	h := a.prov.Host
-	// A new stream is a control plane that knows nothing of where installs
-	// stand, so the next heartbeat tells it again.
-	a.provSent = ""
 	a.mu.Unlock()
+	a.forgetSent()
 	reg.Host = &llmgwv1.Host{
 		OsId: h.OSID, OsVersion: h.OSVersion, OsCodename: h.OSCodename,
 		Arch: h.Arch, CpuFlags: h.CPUFlags, NvidiaDriver: h.NVIDIADriver,
+		// What is measured by now, not what the startup scan found before
+		// the measurement ended: a reconnect would otherwise say 0.
+		MemBandwidthGbs:     a.memBandwidth().All,
+		MemBandwidthCoreGbs: a.memBandwidth().Core,
 	}
 
 	if err := stream.Send(&llmgwv1.AgentMessage{
@@ -606,6 +638,8 @@ func (a *Agent) session(ctx context.Context, server string) error {
 	go a.watchExits(ctx, out, unready(reg.RunningShards))
 	go a.forwardKV(ctx, out)
 	go a.reportCompute(ctx, out)
+	a.setSlurmOut(out)
+	defer a.setSlurmOut(nil)
 
 	for {
 		msg, err := stream.Receive()
@@ -666,6 +700,8 @@ func (a *Agent) heartbeat() *llmgwv1.AgentMessage {
 	// without anybody restarting anything.
 	hb.MemoryTotalMb = int64(discovery.MemoryTotalMB())
 	hb.CpuCount = int32(runtime.NumCPU())
+	bw := a.memBandwidth()
+	hb.MemBandwidthGbs, hb.MemBandwidthCoreGbs = bw.All, bw.Core
 
 	a.reprobeEngines()
 	a.mu.Lock()
@@ -787,6 +823,11 @@ func (a *Agent) sampleLoad() *llmgwv1.NodeLoad {
 				MemoryUsedMb: g.MemoryUsedMB, MemoryTotalMb: g.MemoryTotalMB,
 				SmClockMhz: g.SMClockMHz, EccUncorrected: g.ECCUncorrected,
 			}
+			if p := g.Profile; p != nil {
+				gs.HasProfile, gs.ProfileSource = true, p.Source
+				gs.SmActivePercent, gs.SmOccupancyPercent = p.SMActive, p.SMOccupancy
+				gs.TensorActivePercent, gs.DramActivePercent = p.TensorActive, p.DRAMActive
+			}
 			if w, ok := wear[g.Index]; ok {
 				gs.HasWear = true
 				gs.PcieReplays, gs.RetiredPagesSbe, gs.RetiredPagesDbe, gs.RetirementPending =
@@ -865,10 +906,27 @@ func (a *Agent) removeWeights(paths []string) {
 			a.log.Warn("refused to remove a file llmgw did not fetch", "path", p)
 			continue
 		}
-		if err := os.Remove(abs); err != nil && !os.IsNotExist(err) {
+		// A set — every file of one version, fetched into a directory of its
+		// own — goes whole: os.Remove refused it as not empty, and nothing
+		// was freed of what a model of several files left. Lstat, so a link
+		// is taken away and never what it points at.
+		info, err := os.Lstat(abs)
+		switch {
+		case os.IsNotExist(err):
+			continue
+		case err != nil:
+			a.log.Warn("could not look at cached weights", "path", p, "err", err)
+			continue
+		case info.IsDir():
+			err = os.RemoveAll(abs)
+		default:
+			err = os.Remove(abs)
+		}
+		if err != nil {
 			a.log.Warn("could not remove cached weights", "path", p, "err", err)
 			continue
 		}
+		_ = os.Remove(weights.ChecksumPath(abs))
 		a.log.Info("removed cached weights no model asks for", "path", p)
 	}
 }
@@ -905,6 +963,8 @@ func (a *Agent) handle(ctx context.Context, msg *llmgwv1.ControlMessage, out cha
 		go a.applySettings(m.AgentSettings)
 	case *llmgwv1.ControlMessage_TrainStart:
 		go a.trainJob(ctx, m.TrainStart, out)
+	case *llmgwv1.ControlMessage_Slurm:
+		a.slurmMessage(m.Slurm)
 	case *llmgwv1.ControlMessage_TrainCancel:
 		a.trainer.Cancel(m.TrainCancel.GetJobId())
 	case *llmgwv1.ControlMessage_RestartAgent:
@@ -1120,11 +1180,11 @@ func (a *Agent) startShard(ctx context.Context, m *llmgwv1.StartShard, out chan<
 		send(&llmgwv1.ShardStatus{
 			State: agentsvc.StateToWire(supervisor.Starting),
 			Fetch: &llmgwv1.FetchProgress{
-				Source: p.Source, BytesDone: p.BytesDone, BytesTotal: p.BytesTotal,
+				Source: p.Source, BytesDone: p.BytesDone, BytesTotal: p.BytesTotal, Phase: p.Phase,
 			},
 		})
 	})
-	if path, err := a.weights().Ensure(ctx, spec.Profile.ModelPath); err != nil {
+	if path, err := a.weights().EnsureChecked(ctx, spec.Profile.ModelPath, spec.Profile.AllowPickle, spec.Profile.ModelSHA256); err != nil {
 		stop()
 		report(supervisor.Failed, err.Error(), nil)
 		return
@@ -1143,9 +1203,10 @@ func (a *Agent) startShard(ctx context.Context, m *llmgwv1.StartShard, out chan<
 		for i, m := range ms {
 			stop := a.watchFetch(m["source"], func(p weights.Progress) {
 				send(&llmgwv1.ShardStatus{State: agentsvc.StateToWire(supervisor.Starting),
-					Fetch: &llmgwv1.FetchProgress{Source: p.Source, BytesDone: p.BytesDone, BytesTotal: p.BytesTotal}})
+					Fetch: &llmgwv1.FetchProgress{Source: p.Source, BytesDone: p.BytesDone, BytesTotal: p.BytesTotal,
+						Phase: p.Phase}})
 			})
-			path, err := a.weights().Ensure(ctx, m["source"])
+			path, err := a.weights().Ensure(ctx, m["source"], m["allow_pickle"] == "true")
 			stop()
 			if err != nil {
 				report(supervisor.Failed, fmt.Sprintf("model %s: %v", m["name"], err), nil)
@@ -1161,7 +1222,7 @@ func (a *Agent) startShard(ctx context.Context, m *llmgwv1.StartShard, out chan<
 	// loads them at start, and one missing is a replica that answers for the
 	// base model under the adapter's name.
 	for i, ad := range spec.Profile.Adapters {
-		path, err := a.weights().Ensure(ctx, ad.Path)
+		path, err := a.weights().Ensure(ctx, ad.Path, spec.Profile.AllowPickle)
 		if err != nil {
 			report(supervisor.Failed, fmt.Sprintf("adapter %s: %v", ad.Name, err), nil)
 			return
@@ -1417,11 +1478,14 @@ func Client(cfg tlsconf.Config, timeout time.Duration) (*http.Client, error) {
 // off — is never closed: nothing sends the FIN, and the agent's heartbeats sit
 // in the socket's buffer until the kernel gives up retransmitting, fifteen
 // minutes later. Found by switching off a leader: the standby took over in 14
-// seconds and the node stayed with the dead one. An HTTP/2 ping after ten
-// quiet seconds, given five to answer, finds it in a quarter of a minute.
+// seconds and the node stayed with the dead one. An HTTP/2 ping after five
+// quiet seconds, given three to answer, finds it in eight: by the time the
+// standby leads, as it does within the lease's ten seconds, the node is on
+// its way to it. At fifteen, with the lease at fifteen too, the new leader
+// went without its nodes, and so its judges, for half a minute.
 const (
-	streamIdle = 10 * time.Second
-	streamPing = 5 * time.Second
+	streamIdle = 5 * time.Second
+	streamPing = 3 * time.Second
 )
 
 // localizeSpec fills in what only this node knows.
@@ -1502,18 +1566,25 @@ func ResolveManagement(c Config) (addr, iface string) {
 // shard computes on as soon as that is known: an adopted engine's is found
 // after registration, which carried what was known then.
 func (a *Agent) reportCompute(ctx context.Context, out chan<- *llmgwv1.AgentMessage) {
-	sent := map[string]bool{}
+	// What was sent for each shard, so a compute that changes is sent again:
+	// an adopted shard's kept one is sent at once, and adopting it may then
+	// find it was kept wrong.
+	sent := map[string]string{}
 	t := time.NewTicker(5 * time.Second)
 	defer t.Stop()
 	for range 60 {
 		for _, st := range a.sup.Statuses() {
-			if st.State.Terminal() || sent[st.ID] {
+			if st.State.Terminal() {
 				continue
 			}
 			c := a.sup.ComputeOf(st.ID)
 			if c == nil {
 				continue
 			}
+			said := fmt.Sprint(*c)
+			if sent[st.ID] == said {
+				continue
+			}
 			sh, err := a.sup.Get(st.ID)
 			if err != nil {
 				continue
@@ -1524,7 +1595,7 @@ func (a *Agent) reportCompute(ctx context.Context, out chan<- *llmgwv1.AgentMess
 			}}}
 			select {
 			case out <- msg:
-				sent[st.ID] = true
+				sent[st.ID] = said
 			case <-ctx.Done():
 				return
 			}
@@ -1536,3 +1607,51 @@ func (a *Agent) reportCompute(ctx context.Context, out chan<- *llmgwv1.AgentMess
 		}
 	}
 }
+
+// memBandwidth is the node's memory read rate in GB/s, 0 until measured.
+// Measured in the background, the first time it is asked for: it reads half
+// a gigabyte, which must not hold up an agent reaching its control plane,
+// and the heartbeat carries it when it is known. Not at once but after a
+// minute or so, chosen at random, and again every hour, keeping the best:
+// VMs on one host that all restart together measure each other, and read a
+// third less than any of them gets alone. Once a process rather than an
+// agent, so tests running many agents in one read it once.
+func (a *Agent) memBandwidth() discovery.Bandwidth {
+	bwOnce.Do(func() {
+		go func() {
+			time.Sleep(bandwidthDelay())
+			for {
+				// The best of each seen so far: an hourly pass on a node
+				// busy serving reads less, and that is not the memory.
+				bw := measureBandwidth()
+				cur := a.memBandwidthNow()
+				if bw.All > cur.All || bw.Core > cur.Core {
+					bwBits.Store(math.Float64bits(max(bw.All, cur.All)))
+					coreBits.Store(math.Float64bits(max(bw.Core, cur.Core)))
+					a.log.Info("this node's memory", "read_gbs", math.Round(bw.All*10)/10,
+						"one_core_gbs", math.Round(bw.Core*10)/10)
+				}
+				time.Sleep(time.Hour)
+			}
+		}()
+	})
+	return a.memBandwidthNow()
+}
+
+// memBandwidthNow is what has been measured so far, without starting a
+// measurement.
+func (a *Agent) memBandwidthNow() discovery.Bandwidth {
+	return discovery.Bandwidth{All: math.Float64frombits(bwBits.Load()), Core: math.Float64frombits(coreBits.Load())}
+}
+
+// bandwidthDelay is how long an agent waits before measuring, so agents
+// started together measure apart.
+var bandwidthDelay = func() time.Duration { return 10*time.Second + rand.N(110*time.Second) }
+
+var (
+	bwOnce   sync.Once
+	bwBits   atomic.Uint64
+	coreBits atomic.Uint64
+	// measureBandwidth is replaced in tests, which have no use for it.
+	measureBandwidth = discovery.MeasureMemoryBandwidth
+)

internal/agent/agent_test.go

@@ -68,7 +68,7 @@ func (r *statusRecorder) OnShardCompute(replicaID, shardID string, c reconciler.
 	r.computes = append(r.computes, c)
 }
 
-func (r *statusRecorder) OnFetchProgress(replicaID, shardID, source string, done, total int64) {
+func (r *statusRecorder) OnFetchProgress(replicaID, shardID, source, phase string, done, total int64) {
 	r.mu.Lock()
 	defer r.mu.Unlock()
 	r.fetches = append(r.fetches, fetchReport{replicaID, shardID, source, done, total})
@@ -671,7 +671,7 @@ func TestAShardWithMissingWeightsFailsBeforeLaunching(t *testing.T) {
 	// nobody is watching — so it is caught here, where the message can say
 	// what is actually wrong.
 	a := &Agent{cfg: Config{WeightsCache: t.TempDir()}, log: slog.New(slog.DiscardHandler)}
-	_, err := a.weights().Ensure(t.Context(), filepath.Join(t.TempDir(), "not-there.gguf"))
+	_, err := a.weights().Ensure(t.Context(), filepath.Join(t.TempDir(), "not-there.gguf"), false)
 	if err == nil {
 		t.Fatal("a missing weights file was accepted")
 	}
@@ -925,3 +925,23 @@ func TestABandwidthTestIsRunOnTheAdapterAskedFor(t *testing.T) {
 		t.Errorf("reply %+v, ran %+v", r, got)
 	}
 }
+
+func TestANewStreamIsToldTheFilesAgain(t *testing.T) {
+	// Registration carries what the startup scan found; a file deleted since
+	// came back on a control plane that restarted, because the next
+	// heartbeat saw the list unchanged since it was last sent.
+	dir := t.TempDir()
+	h := start(t, func(c *Config) { c.ModelDirs = []string{dir} })
+	if err := os.WriteFile(filepath.Join(dir, "0123456789abcdef-m.gguf"), []byte("x"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	h.ag.heartbeat()
+	if hb := h.ag.heartbeat().GetHeartbeat(); hb.GetWeightsChanged() {
+		t.Fatal("an unchanged list should not be sent again on the same stream")
+	}
+	h.ag.forgetSent()
+	hb := h.ag.heartbeat().GetHeartbeat()
+	if !hb.GetWeightsChanged() || len(hb.GetWeights()) != 1 {
+		t.Errorf("a new stream should be told the files: changed=%v %v", hb.GetWeightsChanged(), hb.GetWeights())
+	}
+}

internal/agent/discovery/cudatoolkit.go

@@ -0,0 +1,93 @@
+package discovery
+
+import (
+	"encoding/json"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"regexp"
+	"slices"
+	"strconv"
+	"strings"
+)
+
+// CUDAToolkit is the root of an installed CUDA toolkit of this major version
+// — the directory whose bin holds nvcc — or "" when there is none. Engines
+// that compile their kernels where they run (FreeToken, ComfyUI's compiled
+// nodes) need one matching the CUDA their torch was built for, and a
+// toolkit of another major version compiles code that torch cannot load.
+//
+// CUDA_HOME is looked at first, then /usr/local/cuda, then every
+// /usr/local/cuda-<version>, newest first.
+func CUDAToolkit(major int) string {
+	return cudaToolkitIn(major, os.Getenv("CUDA_HOME"), "/usr/local")
+}
+
+func cudaToolkitIn(major int, home, root string) string {
+	var dirs []string
+	if home != "" {
+		dirs = append(dirs, home)
+	}
+	dirs = append(dirs, filepath.Join(root, "cuda"))
+	versioned, _ := filepath.Glob(filepath.Join(root, "cuda-*"))
+	slices.SortFunc(versioned, func(a, b string) int { return versionCompare(b, a) })
+	dirs = append(dirs, versioned...)
+	for _, d := range dirs {
+		if _, err := os.Stat(filepath.Join(d, "bin", "nvcc")); err != nil {
+			continue
+		}
+		if toolkitMajor(d) == major {
+			return d
+		}
+	}
+	return ""
+}
+
+// toolkitMajor is a toolkit's major version, from the version.json every
+// toolkit since 11.1 carries, or from nvcc where it has none; 0 when neither
+// says.
+func toolkitMajor(dir string) int {
+	if b, err := os.ReadFile(filepath.Join(dir, "version.json")); err == nil { //nolint:gosec // a toolkit's own file
+		var v struct {
+			CUDA struct {
+				Version string `json:"version"`
+			} `json:"cuda"`
+		}
+		if json.Unmarshal(b, &v) == nil && v.CUDA.Version != "" {
+			n, _ := strconv.Atoi(strings.SplitN(v.CUDA.Version, ".", 2)[0])
+			return n
+		}
+	}
+	//nolint:gosec // nvcc of a toolkit found on disk
+	out, err := exec.Command(filepath.Join(dir, "bin", "nvcc"), "--version").Output()
+	if err != nil {
+		return 0
+	}
+	if m := nvccRelease.FindSubmatch(out); m != nil {
+		n, _ := strconv.Atoi(string(m[1]))
+		return n
+	}
+	return 0
+}
+
+var nvccRelease = regexp.MustCompile(`release (\d+)\.`)
+
+// versionCompare orders two paths ending in dotted versions: cuda-13.10 is
+// newer than cuda-13.4, which comparing them as text gets wrong.
+func versionCompare(a, b string) int {
+	pa := strings.Split(strings.TrimPrefix(filepath.Base(a), "cuda-"), ".")
+	pb := strings.Split(strings.TrimPrefix(filepath.Base(b), "cuda-"), ".")
+	for i := range max(len(pa), len(pb)) {
+		var x, y int
+		if i < len(pa) {
+			x, _ = strconv.Atoi(pa[i])
+		}
+		if i < len(pb) {
+			y, _ = strconv.Atoi(pb[i])
+		}
+		if x != y {
+			return x - y
+		}
+	}
+	return 0
+}

internal/agent/discovery/cudatoolkit_test.go

@@ -0,0 +1,48 @@
+package discovery
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func toolkit(t *testing.T, root, name, version string) string {
+	t.Helper()
+	d := filepath.Join(root, name)
+	put(t, d, "bin/nvcc", "#!/bin/sh")
+	if version != "" {
+		put(t, d, "version.json", `{"cuda":{"name":"CUDA SDK","version":"`+version+`"}}`)
+	}
+	return d
+}
+
+func TestTheNewestToolkitOfTheMajorVersionAskedForIsFound(t *testing.T) {
+	root := t.TempDir()
+	toolkit(t, root, "cuda-12.8", "12.8.1")
+	toolkit(t, root, "cuda-13.4", "13.4.0")
+	want := toolkit(t, root, "cuda-13.10", "13.10.0")
+	// /usr/local/cuda is a 12: a toolkit of the wrong major version is not
+	// one torch built for 13 can load what it compiles.
+	toolkit(t, root, "cuda", "12.8.1")
+	if got := cudaToolkitIn(13, "", root); got != want {
+		t.Errorf("got %q, want %q", got, want)
+	}
+	if got := cudaToolkitIn(12, "", root); got != filepath.Join(root, "cuda") {
+		t.Errorf("12: got %q", got)
+	}
+	// CUDA_HOME first, when it is one.
+	home := toolkit(t, t.TempDir(), "mine", "13.0.2")
+	if got := cudaToolkitIn(13, home, root); got != home {
+		t.Errorf("CUDA_HOME: got %q", got)
+	}
+	if got := cudaToolkitIn(14, "", root); got != "" {
+		t.Errorf("14: got %q", got)
+	}
+	// A directory without nvcc is not a toolkit.
+	if err := os.MkdirAll(filepath.Join(root, "cuda-13.99"), 0o755); err != nil {
+		t.Fatal(err)
+	}
+	if got := cudaToolkitIn(13, "", root); got != want {
+		t.Errorf("with an empty cuda-13.99: got %q", got)
+	}
+}

internal/agent/discovery/discover.go

@@ -28,6 +28,11 @@ type Options struct {
 	// SkipGPUs omits GPU discovery. A node without nvidia-smi is already
 	// found to have none, so this is for one whose driver is broken.
 	SkipGPUs bool `koanf:"skip_gpus"`
+	// DCGMExporter is the dcgm-exporter metrics URL read for how much of
+	// each card's chip is in use, on cards NVML's GPM does not cover (before
+	// Hopper). Empty is http://127.0.0.1:9400/metrics; "off" reads none. A
+	// node without one costs a refused connection every five minutes.
+	DCGMExporter string `koanf:"dcgm_exporter"`
 	// ModelDirs are scanned for weights. Empty means do not look.
 	ModelDirs []string `koanf:"-"`
 }

internal/agent/discovery/gpm_nonvml.go

@@ -0,0 +1,6 @@
+//go:build !nvml
+
+package discovery
+
+// newGPM has nothing to read GPM with in a build without the nvml tag.
+func newGPM() gpmSource { return nil }

internal/agent/discovery/gpm_nvml.go

@@ -0,0 +1,124 @@
+//go:build nvml
+
+package discovery
+
+import (
+	"sync"
+
+	"github.com/NVIDIA/go-nvml/pkg/nvml"
+)
+
+// nvmlGPM reads GPM, NVML's own profiling counters, on the cards that have
+// them: Hopper and later. Each read takes a sample and works out the metrics
+// against the one before, so NVML stays initialised between reads.
+type nvmlGPM struct {
+	mu     sync.Mutex
+	opened bool
+	failed bool
+	cards  map[int]*gpmCard
+}
+
+type gpmCard struct {
+	dev          nvml.Device
+	prev, cur    nvml.GpmSample
+	havePrevious bool
+}
+
+func newGPM() gpmSource { return &nvmlGPM{} }
+
+// The GPM metrics read, in the order they are asked for.
+var gpmMetrics = []struct {
+	id  nvml.GpmMetricId
+	set func(*GPUProfile, float64)
+}{
+	{nvml.GPM_METRIC_SM_UTIL, func(p *GPUProfile, v float64) { p.SMActive = v }},
+	{nvml.GPM_METRIC_SM_OCCUPANCY, func(p *GPUProfile, v float64) { p.SMOccupancy = v }},
+	{nvml.GPM_METRIC_ANY_TENSOR_UTIL, func(p *GPUProfile, v float64) { p.TensorActive = v }},
+	{nvml.GPM_METRIC_DRAM_BW_UTIL, func(p *GPUProfile, v float64) { p.DRAMActive = v }},
+}
+
+func (g *nvmlGPM) read() map[int]GPUProfile {
+	g.mu.Lock()
+	defer g.mu.Unlock()
+	if g.failed {
+		return nil
+	}
+	if !g.opened && !g.open() {
+		g.failed = true
+		return nil
+	}
+	out := map[int]GPUProfile{}
+	for idx, c := range g.cards {
+		if c.cur.Get(c.dev) != nvml.SUCCESS {
+			// A card in MIG mode, or one that has gone: no sample, and the
+			// next must not be measured against an old one.
+			c.havePrevious = false
+			continue
+		}
+		if c.havePrevious {
+			if p, ok := gpmBetween(c.prev, c.cur); ok {
+				out[idx] = p
+			}
+		}
+		c.prev, c.cur = c.cur, c.prev
+		c.havePrevious = true
+	}
+	return out
+}
+
+// open finds the cards with GPM. False is NVML not starting, or no card
+// having it, and the reader then never tries again.
+func (g *nvmlGPM) open() bool {
+	if nvml.Init() != nvml.SUCCESS {
+		return false
+	}
+	g.cards = map[int]*gpmCard{}
+	n, ret := nvml.DeviceGetCount()
+	if ret != nvml.SUCCESS {
+		n = 0
+	}
+	for i := range n {
+		dev, ret := nvml.DeviceGetHandleByIndex(i)
+		if ret != nvml.SUCCESS {
+			continue
+		}
+		sup, ret := dev.GpmQueryDeviceSupport()
+		if ret != nvml.SUCCESS || sup.IsSupportedDevice == 0 {
+			continue
+		}
+		a, ret := nvml.GpmSampleAlloc()
+		if ret != nvml.SUCCESS {
+			continue
+		}
+		b, ret := nvml.GpmSampleAlloc()
+		if ret != nvml.SUCCESS {
+			_ = a.Free()
+			continue
+		}
+		g.cards[i] = &gpmCard{dev: dev, prev: a, cur: b}
+	}
+	if len(g.cards) == 0 {
+		_ = nvml.Shutdown()
+		return false
+	}
+	g.opened = true
+	return true
+}
+
+// gpmBetween is the profile over the time between two samples.
+func gpmBetween(earlier, later nvml.GpmSample) (GPUProfile, bool) {
+	q := nvml.GpmMetricsGetType{NumMetrics: uint32(len(gpmMetrics)), Sample1: earlier, Sample2: later}
+	for i, m := range gpmMetrics {
+		q.Metrics[i].MetricId = uint32(m.id)
+	}
+	if nvml.GpmMetricsGet(&q) != nvml.SUCCESS {
+		return GPUProfile{}, false
+	}
+	p := GPUProfile{SMActive: -1, SMOccupancy: -1, TensorActive: -1, DRAMActive: -1, Source: "gpm"}
+	for i, m := range gpmMetrics {
+		if r := q.Metrics[i]; r.NvmlReturn == uint32(nvml.SUCCESS) && r.Value >= 0 {
+			m.set(&p, r.Value)
+		}
+	}
+	return p, true
+}

internal/agent/discovery/gpu.go

@@ -52,6 +52,8 @@ type GPU struct {
 	// memory the instance's share. Utilisation stays unreported, because no
 	// driver measures it per instance.
 	MIG *MIGDevice `json:"mig,omitempty"`
+	// PCIe is how the card is linked to its host, zero when not read.
+	PCIe PCIeLink `json:"pcie,omitempty"`
 }
 
 // GPUHealth records conditions that make a GPU unsafe to schedule onto.

internal/agent/discovery/gpu_nvml.go

@@ -67,10 +67,23 @@ func DiscoverGPUsNVML(_ context.Context) ([]GPU, error) {
 		if major, minor, ret := dev.GetCudaComputeCapability(); ret == nvml.SUCCESS {
 			g.ComputeCapability = fmt.Sprintf("%d.%d", major, minor)
 		}
+		if v, ret := dev.GetMaxPcieLinkGeneration(); ret == nvml.SUCCESS {
+			g.PCIe.Gen = v
+		}
+		if v, ret := dev.GetGpuMaxPcieLinkGeneration(); ret == nvml.SUCCESS {
+			g.PCIe.GPUGen = v
+		}
+		if v, ret := dev.GetCurrPcieLinkWidth(); ret == nvml.SUCCESS {
+			g.PCIe.Width = v
+		}
+		if v, ret := dev.GetMaxPcieLinkWidth(); ret == nvml.SUCCESS {
+			g.PCIe.WidthMax = v
+		}
 		if pci, ret := dev.GetPciInfo(); ret == nvml.SUCCESS {
 			g.PCIBusID = busIDString(pci.BusId[:])
 			byBusID[normalizeBusID(g.PCIBusID)] = i
 		}
+		g.PCIe = withSysfs(g.PCIe, "/sys/bus/pci/devices", g.PCIBusID)
 		if t, ret := dev.GetTemperature(nvml.TEMPERATURE_GPU); ret == nvml.SUCCESS {
 			g.Health.TemperatureC = int(t)
 		}

internal/agent/discovery/gpu_smi.go

@@ -64,6 +64,7 @@ func DiscoverGPUsSMI(ctx context.Context, binary string) ([]GPU, error) {
 	if err != nil || len(cards) == 0 {
 		return cards, err
 	}
+	readPCIe(ctx, binary, cards)
 	// The query lists whole cards, MIG mode or not. A card in MIG mode is
 	// replaced by its instances, which only -L names and only the full
 	// report sizes; asked only when some card could be in MIG mode, since

internal/agent/discovery/host.go

@@ -31,6 +31,14 @@ type Host struct {
 	// MinComputeCapability is the oldest GPU's, 0 when there are none or it
 	// is not known. Not read here but from the GPU inventory, by the agent.
 	MinComputeCapability float64 `json:"min_compute_capability,omitempty"`
+	// MemBandwidthGBs is how fast the memory reads, in GB/s, 0 when not
+	// measured: what decides a model's speed on the CPU. Not read here but
+	// measured by the agent when it starts (MeasureMemoryBandwidth).
+	MemBandwidthGBs float64 `json:"mem_bandwidth_gbs,omitempty"`
+	// MemBandwidthCoreGBs is what one core reads alone, in GB/s, 0 when not
+	// measured: a replica given n threads reads at about n times this, up to
+	// MemBandwidthGBs.
+	MemBandwidthCoreGBs float64 `json:"mem_bandwidth_core_gbs,omitempty"`
 }
 
 // Family is the packaging family: "deb", "rpm", or "" when neither is known.

internal/agent/discovery/membw.go

@@ -0,0 +1,169 @@
+package discovery
+
+import (
+	"bufio"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+)
+
+// How fast this machine reads its memory, which on a CPU is what decides how
+// fast a model answers: every token of a reply reads all the weights it uses
+// once, so a node that reads 40 GB/s writes a 4 GB model's tokens about ten
+// times a second, whatever its core count. Measured once when the agent
+// starts, in a fraction of a second, by every CPU summing a share of a buffer
+// too large for any cache — a VM sharing its host's memory with others gets
+// its share, which is the number that matters for it.
+
+// bandwidthBuffer is what is read at most; the least is eight times the
+// processor's last cache, and never under minBuffer. Smaller, it would be
+// the cache that was measured — some hold a gigabyte — and the figure kept
+// would be too high.
+const (
+	bandwidthBuffer = 2 << 30
+	minBuffer       = 256 << 20
+)
+
+// Bandwidth is how fast this machine's memory reads, in GB/s: with every
+// CPU reading, and with one alone. A replica given a few threads of a large
+// machine reads at a few times the second, not at the first: on a server of
+// many memory channels one core cannot keep more than a fraction of them
+// busy.
+type Bandwidth struct {
+	All, Core float64
+}
+
+// coreBuffer is what the one-core passes read: past any cache, and short
+// enough that one core reads it in a fraction of a second.
+const coreBuffer = 512 << 20
+
+// MeasureMemoryBandwidth is the memory's read rate, the best of a few passes
+// each way; zero when the machine has too little free memory to measure.
+func MeasureMemoryBandwidth() Bandwidth {
+	size := int64(bandwidthBuffer)
+	// Never more than a quarter of what is free: a node already holding
+	// models must not be pushed into swap by a measurement. Without room
+	// for a buffer past the cache, none is measured: the hourly retry
+	// measures it when there is.
+	if avail := memAvailable("/proc/meminfo"); avail > 0 {
+		size = min(size, avail/4)
+	}
+	if size < max(minBuffer, 8*lastCacheBytes("/sys/devices/system/cpu")) {
+		return Bandwidth{}
+	}
+	buf := make([]uint64, size/8)
+	for i := range buf {
+		buf[i] = uint64(i)
+	}
+	workers := runtime.NumCPU()
+	chunk := (len(buf) + workers - 1) / workers
+	best := 0.0
+	for pass := 0; pass < 4; pass++ {
+		var wg sync.WaitGroup
+		start := time.Now()
+		for w := 0; w < workers; w++ {
+			lo, hi := w*chunk, min((w+1)*chunk, len(buf))
+			if lo >= hi {
+				continue
+			}
+			wg.Add(1)
+			go func(part []uint64) {
+				defer wg.Done()
+				sink(sum(part))
+			}(buf[lo:hi])
+		}
+		wg.Wait()
+		if s := time.Since(start).Seconds(); s > 0 {
+			best = max(best, float64(size)/s/1e9)
+		}
+	}
+	// And one core alone, over a part of the same buffer.
+	part := buf[:min(len(buf), coreBuffer/8)]
+	core := 0.0
+	for pass := 0; pass < 3; pass++ {
+		start := time.Now()
+		sink(sum(part))
+		if s := time.Since(start).Seconds(); s > 0 {
+			core = max(core, float64(len(part)*8)/s/1e9)
+		}
+	}
+	return Bandwidth{All: best, Core: min(core, best)}
+}
+
+// sum reads every word, eight at a time so the loop is not what limits it.
+func sum(p []uint64) uint64 {
+	var a, b, c, d, e, f, g, h uint64
+	i := 0
+	for ; i+8 <= len(p); i += 8 {
+		a += p[i]
+		b += p[i+1]
+		c += p[i+2]
+		d += p[i+3]
+		e += p[i+4]
+		f += p[i+5]
+		g += p[i+6]
+		h += p[i+7]
+	}
+	for ; i < len(p); i++ {
+		a += p[i]
+	}
+	return a + b + c + d + e + f + g + h
+}
+
+// sinkValue keeps the sums alive, so the reads are not optimised away.
+var sinkValue struct {
+	sync.Mutex
+	v uint64
+}
+
+func sink(v uint64) {
+	sinkValue.Lock()
+	sinkValue.v += v
+	sinkValue.Unlock()
+}
+
+// memAvailable is MemAvailable in bytes, 0 when unknown.
+func memAvailable(path string) int64 {
+	f, err := os.Open(path) //nolint:gosec // a fixed system path
+	if err != nil {
+		return 0
+	}
+	defer f.Close()
+	sc := bufio.NewScanner(f)
+	for sc.Scan() {
+		fields := strings.Fields(sc.Text())
+		if len(fields) >= 2 && fields[0] == "MemAvailable:" {
+			kb, _ := strconv.ParseInt(fields[1], 10, 64)
+			return kb << 10
+		}
+	}
+	return 0
+}
+
+// lastCacheBytes is the largest cache any CPU reports, 0 when unknown.
+func lastCacheBytes(root string) int64 {
+	paths, _ := filepath.Glob(filepath.Join(root, "cpu*", "cache", "index*", "size"))
+	var most int64
+	for _, p := range paths {
+		b, err := os.ReadFile(p) //nolint:gosec // under /sys
+		if err != nil {
+			continue
+		}
+		v := strings.TrimSpace(string(b))
+		mult := int64(1)
+		switch {
+		case strings.HasSuffix(v, "K"):
+			mult, v = 1<<10, strings.TrimSuffix(v, "K")
+		case strings.HasSuffix(v, "M"):
+			mult, v = 1<<20, strings.TrimSuffix(v, "M")
+		}
+		if n, err := strconv.ParseInt(v, 10, 64); err == nil {
+			most = max(most, n*mult)
+		}
+	}
+	return most
+}

internal/agent/discovery/membw_test.go

@@ -0,0 +1,32 @@
+package discovery
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestMemoryBandwidthIsMeasured(t *testing.T) {
+	if testing.Short() {
+		t.Skip("reads half a gigabyte")
+	}
+	bw := MeasureMemoryBandwidth()
+	// Any machine running this reads more than a gigabyte a second and less
+	// than a few terabytes, and one core no more than all of them.
+	if bw.All < 1 || bw.All > 5000 || bw.Core < 0.5 || bw.Core > bw.All {
+		t.Errorf("measured %.1f GB/s, %.1f on one core", bw.All, bw.Core)
+	}
+	t.Logf("%.1f GB/s, %.1f on one core", bw.All, bw.Core)
+}
+
+func TestTheLastCacheIsReadFromSys(t *testing.T) {
+	root := t.TempDir()
+	for path, size := range map[string]string{"cpu0/cache/index0/size": "48K", "cpu0/cache/index3/size": "96M"} {
+		p := filepath.Join(root, path)
+		_ = os.MkdirAll(filepath.Dir(p), 0o755)
+		_ = os.WriteFile(p, []byte(size+"\n"), 0o644)
+	}
+	if got := lastCacheBytes(root); got != 96<<20 {
+		t.Errorf("%d", got)
+	}
+}

internal/agent/discovery/pcie.go

@@ -0,0 +1,181 @@
+package discovery
+
+import (
+	"context"
+	"encoding/csv"
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strconv"
+	"strings"
+)
+
+// A card in the wrong slot: an x16 card in a slot wired x8 or x4, behind a
+// riser, or in a slot of an older PCIe generation than the card's, links
+// slower than it was built to, and nothing says so. A model on the card
+// still decodes at the card's own memory speed, but it loads slower, a
+// tensor-parallel replica over PCIe is slower, and so is anything that
+// spills to the host. The driver says how the link stands and what the
+// card could do; the difference is reported where the node's warnings are.
+
+// PCIeLink is how a card is linked to its host.
+type PCIeLink struct {
+	// Gen is the fastest generation the card and the slot together allow —
+	// not the current one, which drops while the card idles, to save power.
+	// GPUGen is the card's own fastest, 0 when the driver does not say.
+	Gen    int `json:"gen,omitempty"`
+	GPUGen int `json:"gpu_gen,omitempty"`
+	// Width is the lanes in use now, WidthMax the card's most.
+	Width    int `json:"width,omitempty"`
+	WidthMax int `json:"width_max,omitempty"`
+}
+
+// laneGBs is one lane's usable rate in GB/s, by generation.
+var laneGBs = map[int]float64{1: 0.25, 2: 0.5, 3: 0.985, 4: 1.969, 5: 3.938, 6: 7.877}
+
+// GBs is what a link of gen and width carries, in GB/s each way.
+func pcieGBs(gen, width int) float64 { return laneGBs[gen] * float64(width) }
+
+// Problem says how a card's link falls short of what the card could do,
+// empty when it does not or the driver did not say.
+func (p PCIeLink) Problem() string {
+	gen := p.Gen
+	if p.GPUGen > 0 && gen == 0 {
+		gen = p.GPUGen
+	}
+	slowGen := p.GPUGen > 0 && p.Gen > 0 && p.Gen < p.GPUGen
+	narrow := p.WidthMax > 0 && p.Width > 0 && p.Width < p.WidthMax
+	if !slowGen && !narrow {
+		return ""
+	}
+	could := pcieGBs(max(p.GPUGen, gen), p.WidthMax)
+	has := pcieGBs(gen, p.Width)
+	var why []string
+	if narrow {
+		why = append(why, fmt.Sprintf("x%d of its x%d lanes", p.Width, p.WidthMax))
+	}
+	if slowGen {
+		why = append(why, fmt.Sprintf("PCIe Gen%d of its Gen%d", p.Gen, p.GPUGen))
+	}
+	return fmt.Sprintf("linked at %s: %.1f GB/s where it could have %.1f. "+
+		"Check the slot (wired fewer lanes, or an older generation), a riser, or the BIOS's PCIe setting",
+		strings.Join(why, " and "), has, could)
+}
+
+// pcieFields are asked apart from the rest, and the card's own fastest
+// generation apart again: a driver that does not know a field refuses the
+// whole query, and these must not cost a node its cards.
+var pcieFields = [][]string{
+	{"index", "pcie.link.gen.max", "pcie.link.width.current", "pcie.link.width.max", "pcie.link.gen.gpumax"},
+	{"index", "pcie.link.gen.max", "pcie.link.width.current", "pcie.link.width.max"},
+}
+
+// readPCIe fills each card's link from nvidia-smi, leaving it empty when
+// the driver says nothing of it.
+func readPCIe(ctx context.Context, binary string, cards []GPU) {
+	for _, fields := range pcieFields {
+		out, err := exec.CommandContext(ctx, binary, "--query-gpu="+strings.Join(fields, ","),
+			"--format=csv,noheader,nounits").Output()
+		if err != nil {
+			continue
+		}
+		links := parsePCIe(string(out))
+		for i := range cards {
+			if l, ok := links[cards[i].Index]; ok && cards[i].MIG == nil {
+				cards[i].PCIe = l
+			}
+		}
+		break
+	}
+	for i := range cards {
+		if cards[i].MIG == nil {
+			cards[i].PCIe = withSysfs(cards[i].PCIe, "/sys/bus/pci/devices", cards[i].PCIBusID)
+		}
+	}
+}
+
+// withSysfs takes the card's own lane count, and its own fastest
+// generation, from the kernel: the driver's width.max is what card and slot
+// allow together, so a x16 card in a slot wired x8 reads x8 of x8 there,
+// while the device's max_link_width is the card's, 16.
+func withSysfs(l PCIeLink, root, busID string) PCIeLink {
+	dev := sysfsBusID(busID)
+	if dev == "" {
+		return l
+	}
+	read := func(name string) string {
+		b, err := os.ReadFile(filepath.Join(root, dev, name)) //nolint:gosec // under /sys
+		if err != nil {
+			return ""
+		}
+		return strings.TrimSpace(string(b))
+	}
+	if n, err := strconv.Atoi(read("max_link_width")); err == nil && n > 0 {
+		l.WidthMax = n
+	}
+	if n, err := strconv.Atoi(read("current_link_width")); err == nil && n > 0 {
+		l.Width = n
+	}
+	if g := genOfSpeed(read("max_link_speed")); g > 0 && l.GPUGen == 0 {
+		l.GPUGen = g
+	}
+	return l
+}
+
+// sysfsBusID is nvidia-smi's 00000000:3B:00.0 as the kernel names it,
+// 0000:3b:00.0.
+func sysfsBusID(id string) string {
+	id = strings.ToLower(strings.TrimSpace(id))
+	dom, rest, ok := strings.Cut(id, ":")
+	if !ok || len(dom) < 4 {
+		return ""
+	}
+	return dom[len(dom)-4:] + ":" + rest
+}
+
+// genOfSpeed is the PCIe generation of a link speed as sysfs writes it.
+func genOfSpeed(s string) int {
+	switch {
+	case strings.HasPrefix(s, "2.5 GT/s"):
+		return 1
+	case strings.HasPrefix(s, "5.0 GT/s"), strings.HasPrefix(s, "5 GT/s"):
+		return 2
+	case strings.HasPrefix(s, "8.0 GT/s"), strings.HasPrefix(s, "8 GT/s"):
+		return 3
+	case strings.HasPrefix(s, "16.0 GT/s"), strings.HasPrefix(s, "16 GT/s"):
+		return 4
+	case strings.HasPrefix(s, "32.0 GT/s"), strings.HasPrefix(s, "32 GT/s"):
+		return 5
+	case strings.HasPrefix(s, "64.0 GT/s"), strings.HasPrefix(s, "64 GT/s"):
+		return 6
+	}
+	return 0
+}
+
+func parsePCIe(out string) map[int]PCIeLink {
+	r := csv.NewReader(strings.NewReader(out))
+	r.TrimLeadingSpace = true
+	r.FieldsPerRecord = -1
+	rows, _ := r.ReadAll()
+	num := func(s string) int {
+		n, _ := strconv.Atoi(strings.TrimSpace(s))
+		return n
+	}
+	links := map[int]PCIeLink{}
+	for _, row := range rows {
+		if len(row) < 4 {
+			continue
+		}
+		idx, err := strconv.Atoi(strings.TrimSpace(row[0]))
+		if err != nil {
+			continue
+		}
+		l := PCIeLink{Gen: num(row[1]), Width: num(row[2]), WidthMax: num(row[3])}
+		if len(row) > 4 {
+			l.GPUGen = num(row[4])
+		}
+		links[idx] = l
+	}
+	return links
+}

internal/agent/discovery/pcie_test.go

@@ -0,0 +1,54 @@
+package discovery
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestACardInTheWrongSlotIsSaidSo(t *testing.T) {
+	for _, c := range []struct {
+		link PCIeLink
+		want string // empty: nothing to say
+	}{
+		// The PoC's Quadro RTX 6000, idle: current Gen1, but its fastest is 3,
+		// as the card's is, on all 16 lanes. Nothing wrong.
+		{PCIeLink{Gen: 3, GPUGen: 3, Width: 16, WidthMax: 16}, ""},
+		// An x16 card in a slot wired x8.
+		{PCIeLink{Gen: 4, GPUGen: 4, Width: 8, WidthMax: 16}, "x8 of its x16 lanes: 15.8 GB/s where it could have 31.5"},
+		// A Gen5 card in a Gen4 slot.
+		{PCIeLink{Gen: 4, GPUGen: 5, Width: 16, WidthMax: 16}, "PCIe Gen4 of its Gen5: 31.5 GB/s where it could have 63.0"},
+		// Both, behind a riser.
+		{PCIeLink{Gen: 3, GPUGen: 4, Width: 4, WidthMax: 16}, "x4 of its x16 lanes and PCIe Gen3 of its Gen4"},
+		// A driver that says nothing.
+		{PCIeLink{}, ""},
+	} {
+		got := c.link.Problem()
+		if (c.want == "") != (got == "") || !strings.Contains(got, c.want) {
+			t.Errorf("%+v: %q, want %q", c.link, got, c.want)
+		}
+	}
+}
+
+func TestTheLinkIsReadFromNvidiaSMI(t *testing.T) {
+	links := parsePCIe("0, 3, 16, 16, 3\n1, 4, 8, 16, [N/A]\n")
+	if links[0] != (PCIeLink{Gen: 3, GPUGen: 3, Width: 16, WidthMax: 16}) || links[1] != (PCIeLink{Gen: 4, Width: 8, WidthMax: 16}) {
+		t.Errorf("%+v", links)
+	}
+}
+
+func TestTheCardsOwnLanesComeFromTheKernel(t *testing.T) {
+	// The driver says x8 of x8 for a x16 card in a slot wired x8; the
+	// device's own max_link_width says 16.
+	root := t.TempDir()
+	dev := filepath.Join(root, "0000:3b:00.0")
+	_ = os.MkdirAll(dev, 0o755)
+	for f, v := range map[string]string{"max_link_width": "16", "current_link_width": "8", "max_link_speed": "16.0 GT/s PCIe"} {
+		_ = os.WriteFile(filepath.Join(dev, f), []byte(v+"\n"), 0o644)
+	}
+	l := withSysfs(PCIeLink{Gen: 4, Width: 8, WidthMax: 8}, root, "00000000:3B:00.0")
+	if l.WidthMax != 16 || l.Width != 8 || l.GPUGen != 4 || !strings.Contains(l.Problem(), "x8 of its x16") {
+		t.Errorf("%+v: %q", l, l.Problem())
+	}
+}

internal/agent/discovery/prof.go

@@ -0,0 +1,176 @@
+package discovery
+
+import (
+	"context"
+	"io"
+	"net/http"
+	"strconv"
+	"strings"
+	"time"
+)
+
+// GPUProfile is how much of a card's chip the work on it uses. nvidia-smi's
+// utilization.gpu cannot say: it is the share of time any kernel ran, and a
+// kernel keeping one SM of a hundred busy reads 100%. Measured on a Quadro
+// RTX 6000 under decode, it read 92% while the SMs had work 69% of the time
+// and issued an instruction on 18% of their cycles.
+//
+// These come from the card's profiling counters, which only two things read
+// without stopping the work: NVML's GPM, on Hopper and later, and DCGM, which
+// a site may run as dcgm-exporter. Percentages; -1 is not reported.
+type GPUProfile struct {
+	// SMActive is the share of time the SMs had work, averaged over all of
+	// them; SMOccupancy how full those SMs were with warps.
+	SMActive, SMOccupancy float64
+	// TensorActive is the tensor cores. Decode at batch one hardly uses them
+	// even when nothing is wrong: a quantised matrix-vector product runs on
+	// the integer units. Prefill and batches are where they show.
+	TensorActive float64
+	// DRAMActive is the memory interface, which decode is bound by.
+	DRAMActive float64
+	// Source is "gpm" or "dcgm".
+	Source string
+}
+
+// DefaultDCGMExporter is where dcgm-exporter listens unless told otherwise.
+const DefaultDCGMExporter = "http://127.0.0.1:9400/metrics"
+
+// dcgmRetry is how long a dcgm-exporter that did not answer is left alone.
+// Most nodes run none, and asking every ten seconds would be asking nobody.
+const dcgmRetry = 5 * time.Minute
+
+// gpmSource reads GPM, where this build and the card have it.
+type gpmSource interface {
+	read() map[int]GPUProfile
+}
+
+// ProfSampler reads each card's GPUProfile: from GPM where the card has it,
+// otherwise from a dcgm-exporter on the node.
+type ProfSampler struct {
+	gpm  gpmSource
+	dcgm string
+	get  func(ctx context.Context, url string) ([]byte, error)
+	now  func() time.Time
+
+	dcgmAfter time.Time
+}
+
+// NewProfSampler reads this machine. dcgmURL is the dcgm-exporter to ask:
+// empty is DefaultDCGMExporter, "off" asks none.
+func NewProfSampler(dcgmURL string) *ProfSampler {
+	switch dcgmURL {
+	case "":
+		dcgmURL = DefaultDCGMExporter
+	case "off":
+		dcgmURL = ""
+	}
+	return &ProfSampler{gpm: newGPM(), dcgm: dcgmURL, get: httpGet, now: time.Now}
+}
+
+// Sample is each card's profile by index; a card missing has none. GPM needs
+// two samples, so its cards appear from the second call.
+func (p *ProfSampler) Sample(ctx context.Context) map[int]GPUProfile {
+	out := map[int]GPUProfile{}
+	if p.gpm != nil {
+		for i, v := range p.gpm.read() {
+			out[i] = v
+		}
+	}
+	if p.dcgm == "" || p.now().Before(p.dcgmAfter) {
+		return out
+	}
+	body, err := p.get(ctx, p.dcgm)
+	if err != nil {
+		p.dcgmAfter = p.now().Add(dcgmRetry)
+		return out
+	}
+	for i, v := range parseDCGM(body) {
+		if _, ok := out[i]; !ok {
+			out[i] = v
+		}
+	}
+	return out
+}
+
+func httpGet(ctx context.Context, url string) ([]byte, error) {
+	ctx, cancel := context.WithTimeout(ctx, 2*time.Second)
+	defer cancel()
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return nil, err
+	}
+	resp, err := http.DefaultClient.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode != http.StatusOK {
+		return nil, io.ErrUnexpectedEOF
+	}
+	return io.ReadAll(io.LimitReader(resp.Body, 4<<20))
+}
+
+// The dcgm-exporter fields read, as ratios of one. Its default field list
+// has them on every card DCGM can profile.
+var dcgmFields = map[string]func(*GPUProfile, float64){
+	"DCGM_FI_PROF_SM_ACTIVE":          func(p *GPUProfile, v float64) { p.SMActive = v },
+	"DCGM_FI_PROF_SM_OCCUPANCY":       func(p *GPUProfile, v float64) { p.SMOccupancy = v },
+	"DCGM_FI_PROF_PIPE_TENSOR_ACTIVE": func(p *GPUProfile, v float64) { p.TensorActive = v },
+	"DCGM_FI_PROF_DRAM_ACTIVE":        func(p *GPUProfile, v float64) { p.DRAMActive = v },
+}
+
+// parseDCGM reads dcgm-exporter's text exposition: a line per field and
+// card, the card in the gpu label. A card with none of the fields — DCGM
+// cannot profile it — is left out. A MIG instance's lines carry a
+// GPU_I_ID and are left out too: these are whole cards.
+func parseDCGM(body []byte) map[int]GPUProfile {
+	out := map[int]GPUProfile{}
+	for _, line := range strings.Split(string(body), "\n") {
+		if line == "" || line[0] == '#' {
+			continue
+		}
+		open := strings.IndexByte(line, '{')
+		end := strings.LastIndexByte(line, '}')
+		if open < 0 || end < open {
+			continue
+		}
+		set, ok := dcgmFields[line[:open]]
+		if !ok {
+			continue
+		}
+		labels := line[open+1 : end]
+		if strings.Contains(labels, `GPU_I_ID="`) && !strings.Contains(labels, `GPU_I_ID=""`) {
+			continue
+		}
+		idx, err := strconv.Atoi(labelValue(labels, "gpu"))
+		if err != nil {
+			continue
+		}
+		val := strings.Fields(line[end+1:])
+		if len(val) == 0 {
+			continue
+		}
+		v, err := strconv.ParseFloat(val[0], 64)
+		if err != nil || v < 0 {
+			continue
+		}
+		p, seen := out[idx]
+		if !seen {
+			p = GPUProfile{SMActive: -1, SMOccupancy: -1, TensorActive: -1, DRAMActive: -1, Source: "dcgm"}
+		}
+		set(&p, 100*v)
+		out[idx] = p
+	}
+	return out
+}
+
+// labelValue is one label's value from a Prometheus label set.
+func labelValue(labels, name string) string {
+	for _, kv := range strings.Split(labels, ",") {
+		k, v, ok := strings.Cut(strings.TrimSpace(kv), "=")
+		if ok && k == name {
+			return strings.Trim(v, `"`)
+		}
+	}
+	return ""
+}

internal/agent/discovery/prof_test.go

@@ -0,0 +1,107 @@
+package discovery
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+)
+
+// dcgmBody is dcgm-exporter's output for two cards and a MIG instance:
+// card 0 profiled, card 1 a model DCGM cannot profile, which has the device
+// fields and none of the PROF ones.
+const dcgmBody = `# HELP DCGM_FI_DEV_GPU_UTIL GPU utilization (in %).
+# TYPE DCGM_FI_DEV_GPU_UTIL gauge
+DCGM_FI_DEV_GPU_UTIL{gpu="0",UUID="GPU-a",device="nvidia0",modelName="NVIDIA L40S",Hostname="n1"} 100
+DCGM_FI_DEV_GPU_UTIL{gpu="1",UUID="GPU-b",device="nvidia1",modelName="Quadro RTX 6000",Hostname="n1"} 92
+# HELP DCGM_FI_PROF_SM_ACTIVE The ratio of cycles an SM has at least 1 warp assigned.
+# TYPE DCGM_FI_PROF_SM_ACTIVE gauge
+DCGM_FI_PROF_SM_ACTIVE{gpu="0",UUID="GPU-a",device="nvidia0",modelName="NVIDIA L40S",Hostname="n1"} 0.007042
+DCGM_FI_PROF_SM_ACTIVE{gpu="2",UUID="GPU-c",device="nvidia2",modelName="NVIDIA A100",GPU_I_PROFILE="3g.40gb",GPU_I_ID="1",Hostname="n1"} 0.5
+DCGM_FI_PROF_SM_OCCUPANCY{gpu="0",UUID="GPU-a",device="nvidia0",modelName="NVIDIA L40S",Hostname="n1"} 0.25
+DCGM_FI_PROF_PIPE_TENSOR_ACTIVE{gpu="0",UUID="GPU-a",device="nvidia0",modelName="NVIDIA L40S",Hostname="n1"} 0
+`
+
+func TestDCGMExporterSaysHowMuchOfEachChipIsInUse(t *testing.T) {
+	got := parseDCGM([]byte(dcgmBody))
+	if len(got) != 1 {
+		t.Fatalf("want card 0 alone (card 1 unprofiled, card 2 a MIG instance): %+v", got)
+	}
+	p := got[0]
+	// The LinkedIn L40S: 100% busy, one SM of 142 at work.
+	if p.SMActive < 0.70 || p.SMActive > 0.71 || p.SMOccupancy != 25 || p.TensorActive != 0 || p.Source != "dcgm" {
+		t.Errorf("card 0 %+v", p)
+	}
+	// A field the exporter does not send is unknown, not zero.
+	if p.DRAMActive != -1 {
+		t.Errorf("DRAM active not sent, read as %v", p.DRAMActive)
+	}
+}
+
+type fakeGPM map[int]GPUProfile
+
+func (f fakeGPM) read() map[int]GPUProfile { return f }
+
+func TestGPMIsPreferredAndDCGMFillsTheCardsItDoesNotCover(t *testing.T) {
+	calls := 0
+	p := &ProfSampler{
+		gpm:  fakeGPM{1: {SMActive: 40, Source: "gpm"}},
+		dcgm: DefaultDCGMExporter,
+		get: func(context.Context, string) ([]byte, error) {
+			calls++
+			return []byte(`DCGM_FI_PROF_SM_ACTIVE{gpu="0"} 0.5` + "\n" + `DCGM_FI_PROF_SM_ACTIVE{gpu="1"} 0.9` + "\n"), nil
+		},
+		now: time.Now,
+	}
+	got := p.Sample(t.Context())
+	if got[0].Source != "dcgm" || got[0].SMActive != 50 {
+		t.Errorf("card 0 %+v", got[0])
+	}
+	if got[1].Source != "gpm" || got[1].SMActive != 40 {
+		t.Errorf("GPM's card was taken from DCGM: %+v", got[1])
+	}
+	if calls != 1 {
+		t.Errorf("%d calls", calls)
+	}
+}
+
+func TestANodeWithoutDCGMIsNotAskedEverySample(t *testing.T) {
+	now := time.Now()
+	calls := 0
+	p := &ProfSampler{dcgm: DefaultDCGMExporter, now: func() time.Time { return now },
+		get: func(context.Context, string) ([]byte, error) { calls++; return nil, errors.New("connection refused") }}
+	for range 5 {
+		if got := p.Sample(t.Context()); len(got) != 0 {
+			t.Fatalf("%+v", got)
+		}
+		now = now.Add(10 * time.Second)
+	}
+	if calls != 1 {
+		t.Errorf("asked %d times in 50 s", calls)
+	}
+	now = now.Add(dcgmRetry)
+	p.Sample(t.Context())
+	if calls != 2 {
+		t.Errorf("not asked again after %v", dcgmRetry)
+	}
+	if off := NewProfSampler("off"); off.dcgm != "" {
+		t.Errorf("off still asks %q", off.dcgm)
+	}
+}
+
+func TestEachSampleCarriesTheProfileOfTheCardsThatHaveOne(t *testing.T) {
+	smi := func(context.Context, string) ([]byte, error) {
+		return []byte("0, 71, 239, 260, 0x0, 92, 30, 23000, 24576, 1878, 0, 0\n" +
+			"1, 30, 20, 260, 0x0, 0, 0, 512, 24576, 300, 0, 0\n"), nil
+	}
+	now := time.Now()
+	s := sampler(t.TempDir(), smi, &now).WithProfile(&ProfSampler{
+		gpm: fakeGPM{0: {SMActive: 69, SMOccupancy: 45, TensorActive: 1, DRAMActive: 33, Source: "gpm"}}, now: time.Now})
+	got := s.Sample(t.Context()).GPUs
+	if got[0].Profile == nil || got[0].Profile.SMActive != 69 || got[0].UtilPercent != 92 {
+		t.Errorf("card 0 %+v", got[0])
+	}
+	if got[1].Profile != nil {
+		t.Errorf("card 1 has a profile nobody read: %+v", got[1].Profile)
+	}
+}

internal/agent/discovery/sensors.go

@@ -49,6 +49,9 @@ type GPUSensor struct {
 	MemoryUsedMB, MemoryTotalMB float64
 	SMClockMHz                  float64
 	ECCUncorrected              float64
+	// Profile is how much of the chip the work uses, where something on the
+	// node can say: nil on most cards. UtilPercent is not that.
+	Profile *GPUProfile
 }
 
 // sensorEvery bounds how often nvidia-smi is run. Heartbeats come more often
@@ -61,6 +64,8 @@ type SensorSampler struct {
 	sys string // /sys, or a test's copy of it
 	smi func(ctx context.Context, fields string) ([]byte, error)
 	now func() time.Time
+	// prof reads the cards' profiling counters; nil reads none.
+	prof *ProfSampler
 
 	mu         sync.Mutex
 	last       Sensors
@@ -74,6 +79,19 @@ func NewSensorSampler() *SensorSampler {
 	return &SensorSampler{sys: "/sys", smi: runSMI, now: time.Now}
 }
 
+// WithProfile has each sample carry the cards' profiles as well.
+func (s *SensorSampler) WithProfile(p *ProfSampler) *SensorSampler {
+	s.prof = p
+	return s
+}
+
+// SampleGPUs reads every card's heat, power and how busy it is now, without
+// a sampler's ten seconds of memory: for a screen that asks every few seconds
+// and wants what it asked for. Nil on a machine without nvidia-smi or a card.
+func SampleGPUs(ctx context.Context) []GPUSensor {
+	return (&SensorSampler{smi: runSMI}).gpus(ctx)
+}
+
 func runSMI(ctx context.Context, fields string) ([]byte, error) {
 	ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
 	defer cancel()
@@ -94,6 +112,14 @@ func (s *SensorSampler) Sample(ctx context.Context) Sensors {
 	out.CPUTempC, out.InletTempC = s.hwmon()
 	out.CPUPowerW = s.rapl(now)
 	out.GPUs = s.gpus(ctx)
+	if s.prof != nil {
+		profs := s.prof.Sample(ctx)
+		for i := range out.GPUs {
+			if p, ok := profs[out.GPUs[i].Index]; ok {
+				out.GPUs[i].Profile = &p
+			}
+		}
+	}
 	s.last, s.lastAt = out, now
 	return out
 }

internal/agent/engines/engine.go

@@ -47,6 +47,13 @@ type Profile struct {
 	// KVEvents has the engine publish its KV cache events to the agent,
 	// for the router's precise index. vLLM only.
 	KVEvents bool `json:"kv_events,omitempty"`
+	// AllowPickle lets the node fetch weights in a pickle format, which run
+	// code when loaded; the catalog's allow_pickle.
+	AllowPickle bool `json:"allow_pickle,omitempty"`
+	// ModelSHA256 is the sha256 the registry published for the file
+	// ModelPath names; the node refuses a copy that is not it. Empty when
+	// nothing is known to check against.
+	ModelSHA256 string `json:"model_sha256,omitempty"`
 }
 
 // ShardSpec is one shard of one replica on this node.
@@ -215,6 +222,8 @@ func NewRegistry(cfg Config) *Registry {
 	r.Add(&TRTLLM{cfg: cfg.TRTLLM})
 	r.Add(&Sim{cfg: cfg.Sim})
 	r.Add(&ComfyUI{cfg: cfg.ComfyUI})
+	r.Add(&Strata{cfg: cfg.Strata})
+	r.Add(&FreeToken{cfg: cfg.FreeToken})
 	return r
 }
 
@@ -265,6 +274,13 @@ type Config struct {
 	Sim    SimConfig    `koanf:"sim"`
 	// ComfyUI makes pictures; offered by a node whose ComfyUI is named here.
 	ComfyUI ComfyUIConfig `koanf:"comfyui"`
+	// Strata serves a few mixture-of-experts models from RAM and one card;
+	// offered by a node whose Strata is named here, or that installed it.
+	Strata StrataConfig `koanf:"strata"`
+	// FreeToken serves mixture-of-experts models larger than a card from
+	// RAM and one card of 8.0 or newer; installed by the agent where the
+	// cards allow it, or named here.
+	FreeToken FreeTokenConfig `koanf:"freetoken"`
 }
 
 // --- shared helpers ---

internal/agent/engines/engines_test.go

@@ -477,14 +477,14 @@ func TestValidateRequiresModelPath(t *testing.T) {
 
 func TestRegistry(t *testing.T) {
 	r := NewRegistry(Config{})
-	if got := r.Names(); !slices.Equal(got, []string{"comfyui", "llamacpp", "sglang", "sim", "trtllm", "vllm"}) {
+	if got := r.Names(); !slices.Equal(got, []string{"comfyui", "freetoken", "llamacpp", "sglang", "sim", "strata", "trtllm", "vllm"}) {
 		t.Errorf("registry contents: %v", got)
 	}
 	if _, err := r.Get("vllm"); err != nil {
 		t.Errorf("vllm missing: %v", err)
 	}
 	err := func() error { _, err := r.Get("nope"); return err }()
-	if err == nil || !strings.Contains(err.Error(), "llamacpp, sglang, sim, trtllm, vllm") {
+	if err == nil || !strings.Contains(err.Error(), "llamacpp, sglang, sim, strata, trtllm, vllm") {
 		t.Errorf("an unknown engine should list what is available: %v", err)
 	}
 }

internal/agent/engines/freetoken.go

@@ -0,0 +1,137 @@
+package engines
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"sync/atomic"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
+)
+
+// FreeTokenConfig locates a FreeToken installation.
+type FreeTokenConfig struct {
+	// Binary is its ft command, in the environment it was installed into.
+	// The agent installs one itself on a node whose GPUs it runs on; this
+	// is for one installed by hand.
+	Binary string `koanf:"binary"`
+}
+
+// freetokenToolkit finds the CUDA 13 toolkit; a variable for tests.
+var freetokenToolkit = func() string { return discovery.CUDAToolkit(13) }
+
+// FreeToken runs FreeToken (github.com/FlashML-org/FreeToken): a
+// mixture-of-experts engine that keeps a model's experts in host RAM, the
+// ones used most in VRAM, and streams the rest over PCIe, so a model several
+// times a card's size serves from one card. It reads Hugging Face
+// safetensors checkpoints as they are published (FP8, NVFP4, MXFP4, BF16)
+// and serves the OpenAI and Anthropic APIs. One card of compute capability
+// 8.0 or newer; it compiles some of its kernels on first start, with the
+// CUDA 13 toolkit.
+type FreeToken struct {
+	cfg       FreeTokenConfig
+	installed atomic.Pointer[string]
+}
+
+// Name identifies the engine in the catalog.
+func (*FreeToken) Name() string { return "freetoken" }
+
+// HealthPath answers before the model is loaded too, with its progress in
+// the body; the supervisor waits for the status code, so it counts as ready
+// only once FreeToken's own startup — loading, warm-up, CUDA graphs — is
+// over and the API server is accepting.
+func (*FreeToken) HealthPath() string { return "/health" }
+
+// MetricsPath is empty: FreeToken serves its numbers as JSON (/v1/stats),
+// not in Prometheus' text, so the router counts its requests itself.
+func (*FreeToken) MetricsPath() string { return "" }
+
+// ServesHTTP is true: a FreeToken replica is one process on one node.
+func (*FreeToken) ServesHTTP(ShardSpec) bool { return true }
+
+// UseInstalled points the engine at the ft the agent installed.
+func (f *FreeToken) UseInstalled(path string) { f.installed.Store(&path) }
+
+// Configured reports whether the configuration names a FreeToken.
+func (f *FreeToken) Configured() bool { return f.cfg.Binary != "" }
+
+func (f *FreeToken) binary() string {
+	if p := f.installed.Load(); p != nil {
+		return *p
+	}
+	return f.cfg.Binary
+}
+
+// HelpCommand is ft serve's own help.
+func (f *FreeToken) HelpCommand(Command) (string, []string) {
+	return f.binary(), []string{"serve", "--help"}
+}
+
+// Probe reports the FreeToken a configuration names, by what it says its
+// version is.
+func (f *FreeToken) Probe(ctx context.Context) (string, error) {
+	bin := f.binary()
+	if bin == "" {
+		return "", fmt.Errorf("no FreeToken is configured")
+	}
+	out, err := exec.CommandContext(ctx, bin, "--version").Output() //nolint:gosec // the configured engine
+	if err != nil {
+		return "", fmt.Errorf("%s does not run: %w", bin, err)
+	}
+	// "freetoken version 0.2.1"
+	if fs := strings.Fields(string(out)); len(fs) > 0 {
+		return fs[len(fs)-1], nil
+	}
+	return Unknown, nil
+}
+
+// Validate checks the profile: one card of 8.0 or newer, on one node.
+// FreeToken has no tensor parallel; it picks one device.
+func (*FreeToken) Validate(p Profile, gpus []GPUInfo) error {
+	if p.Nodes > 1 || p.TP > 1 || p.PP > 1 || p.GPUsPerNode > 1 {
+		return fmt.Errorf("profile %q: freetoken runs on one GPU of one node", p.Name)
+	}
+	if err := requireComputeCapability(p, gpus, 8.0, "freetoken"); err != nil {
+		return err
+	}
+	return validateCommon(p, gpus)
+}
+
+// Commands is ft serve. The card is the shard's, by CUDA_VISIBLE_DEVICES,
+// so it is device 0 to FreeToken; its key goes in the environment, where it
+// reads it from, rather than on a command line every user can read.
+func (f *FreeToken) Commands(sh ShardSpec) ([]Command, error) {
+	bin := f.binary()
+	if bin == "" {
+		return nil, fmt.Errorf("freetoken: no FreeToken is installed or configured (engines.freetoken.binary)")
+	}
+	toolkit := freetokenToolkit()
+	if toolkit == "" {
+		return nil, fmt.Errorf("freetoken compiles its kernels on first start and this node has no CUDA 13 toolkit " +
+			"(nvcc in CUDA_HOME, /usr/local/cuda or /usr/local/cuda-13.*)")
+	}
+	p := sh.Profile
+	args := []string{"serve",
+		"--model", p.ModelPath,
+		"--host", sh.Host(),
+		"--port", strconv.Itoa(sh.HTTPPort),
+		"--gpu", "0",
+	}
+	if p.ServedModelName != "" {
+		args = append(args, "--served-model-name", p.ServedModelName)
+	}
+	args = append(args, p.Args...)
+	env := baseEnv(sh)
+	env["CUDA_HOME"] = toolkit
+	// nvcc for the kernels it compiles, and the environment's own bin
+	// first, where ninja is.
+	env["PATH"] = strings.Join([]string{filepath.Dir(bin), filepath.Join(toolkit, "bin"), os.Getenv("PATH")}, ":")
+	if sh.APIKey != "" {
+		env["FREETOKEN_API_KEY"] = sh.APIKey
+	}
+	return []Command{{Name: "freetoken", Path: bin, Args: args, Env: materializeEnv(env)}}, nil
+}

internal/agent/engines/freetoken_test.go

@@ -0,0 +1,112 @@
+package engines
+
+import (
+	"slices"
+	"strings"
+	"testing"
+)
+
+func withToolkit(t *testing.T, dir string) {
+	t.Helper()
+	old := freetokenToolkit
+	freetokenToolkit = func() string { return dir }
+	t.Cleanup(func() { freetokenToolkit = old })
+}
+
+func TestFreeTokenServesOnTheShardsCardWithTheKeyInTheEnvironment(t *testing.T) {
+	withToolkit(t, "/usr/local/cuda-13.4")
+	f := &FreeToken{}
+	f.UseInstalled("/opt/llmgw/engines/freetoken/0.2.1-cu13/bin/ft")
+	spec := ShardSpec{
+		ReplicaID: "r1", HTTPPort: 18091, BindAddr: "10.0.0.160", GPUIndices: []int{3},
+		APIKey: "secret-key",
+		Profile: Profile{Name: "p", Engine: "freetoken", TP: 1, ModelPath: "/var/lib/llmgw-weights/glm-5.2-nvfp4",
+			ServedModelName: "glm-5.2", Args: []string{"--max-running-requests", "8"}},
+	}
+	cmds, err := f.Commands(spec)
+	if err != nil {
+		t.Fatal(err)
+	}
+	c := cmds[0]
+	if c.Path != "/opt/llmgw/engines/freetoken/0.2.1-cu13/bin/ft" {
+		t.Errorf("path %s", c.Path)
+	}
+	args := strings.Join(c.Args, " ")
+	for _, want := range []string{
+		"serve --model /var/lib/llmgw-weights/glm-5.2-nvfp4", "--host 10.0.0.160", "--port 18091",
+		// The shard's card is device 0 once CUDA_VISIBLE_DEVICES pins it.
+		"--gpu 0", "--served-model-name glm-5.2", "--max-running-requests 8",
+	} {
+		if !strings.Contains(args, want) {
+			t.Errorf("args lack %q: %s", want, args)
+		}
+	}
+	if strings.Contains(args, "secret-key") {
+		t.Error("the key is on the command line, which every user on the node can read")
+	}
+	for _, want := range []string{"FREETOKEN_API_KEY=secret-key", "CUDA_VISIBLE_DEVICES=3", "CUDA_HOME=/usr/local/cuda-13.4"} {
+		if !slices.Contains(c.Env, want) {
+			t.Errorf("env lacks %s", want)
+		}
+	}
+	var path string
+	for _, e := range c.Env {
+		if p, ok := strings.CutPrefix(e, "PATH="); ok {
+			path = p
+		}
+	}
+	if !strings.HasPrefix(path, "/opt/llmgw/engines/freetoken/0.2.1-cu13/bin:/usr/local/cuda-13.4/bin:") {
+		t.Errorf("nvcc and the environment's tools are not first on PATH: %s", path)
+	}
+}
+
+func TestFreeTokenWithoutTheToolkitSaysSoInsteadOfFailingOnItsFirstModel(t *testing.T) {
+	withToolkit(t, "")
+	f := &FreeToken{cfg: FreeTokenConfig{Binary: "/x/bin/ft"}}
+	_, err := f.Commands(ShardSpec{HTTPPort: 1, Profile: Profile{Name: "p", ModelPath: "/m"}})
+	if err == nil || !strings.Contains(err.Error(), "CUDA 13 toolkit") {
+		t.Errorf("got %v", err)
+	}
+}
+
+func TestFreeTokenRunsOnOneAmpereOrNewerCard(t *testing.T) {
+	f := &FreeToken{}
+	p := Profile{Name: "p", Engine: "freetoken", TP: 1, ModelPath: "/m"}
+	if err := f.Validate(p, []GPUInfo{{Name: "NVIDIA RTX A6000", ComputeCapability: "8.6", MemoryTotalMB: 49140}}); err != nil {
+		t.Errorf("an Ampere card was refused: %v", err)
+	}
+	// Tried on cuda01: no kernel image for the card, then an illegal memory
+	// access in fp16.
+	if err := f.Validate(p, []GPUInfo{{Name: "Quadro RTX 6000", ComputeCapability: "7.5", MemoryTotalMB: 24576}}); err == nil {
+		t.Error("a Turing card was accepted")
+	}
+	p.TP = 2
+	if err := f.Validate(p, h100(2)); err == nil {
+		t.Error("a two-card profile was accepted; FreeToken has no tensor parallel")
+	}
+}
+
+func TestFreeTokenFlagsThatReadFilesOrMoveWhatTheAgentSetsAreRefused(t *testing.T) {
+	for _, a := range []string{
+		"--gpu", "--served-model-name=x", "--mmproj", "--allowed-local-media-path",
+		// argparse takes a beginning for the whole flag.
+		"--allowed-local", "--api-key", "--port", "--hf-overrides",
+	} {
+		if UnsafeArgs("freetoken", []string{a, "v"}) == nil {
+			t.Errorf("%s was allowed", a)
+		}
+	}
+	for _, a := range []string{"--max-running-requests", "--moe-strategy", "--dtype", "--max-seq-len-override", "--moe-cache-auto"} {
+		if err := UnsafeArgs("freetoken", []string{a, "1"}); err != nil {
+			t.Errorf("%s was refused: %v", a, err)
+		}
+	}
+	for k, refuse := range map[string]bool{
+		"FREETOKEN_PIN_BUDGET_GB": false, "FREETOKEN_API_KEY": true, "FREETOKEN_API_LOG_DIR": true,
+		"FREETOKEN_MMPROJ": true, "FREETOKEN_HOST": true,
+	} {
+		if got, why := UnsafeEnv(k); got != refuse {
+			t.Errorf("%s: refused %v (%s), want %v", k, got, why, refuse)
+		}
+	}
+}

internal/agent/engines/pickle_test.go

@@ -0,0 +1,58 @@
+package engines
+
+import (
+	"slices"
+	"testing"
+)
+
+func TestAModelDownloadedByTheEngineIsReadAsSafetensorsOnly(t *testing.T) {
+	for path, want := range map[string]bool{
+		"Qwen/Qwen2.5-7B-Instruct":    true,
+		"/var/lib/llmgw/weights/abc/": false, "./local": false,
+		"http://m/q/v/": false, "q.gguf": false,
+	} {
+		got := slices.Contains(safetensorsOnly(Profile{ModelPath: path}), "safetensors")
+		if got != want {
+			t.Errorf("%q: safetensors only = %v, want %v", path, got, want)
+		}
+	}
+	if len(safetensorsOnly(Profile{ModelPath: "Qwen/Qwen2.5-7B-Instruct", AllowPickle: true})) != 0 {
+		t.Error("allow_pickle still forced safetensors")
+	}
+}
+
+func TestAForbiddenFlagIsKnownHoweverItIsSpelled(t *testing.T) {
+	for engine, args := range map[string][]string{
+		"vllm":   {"--trust_remote_code"},
+		"sglang": {"--load-format", "pt"},
+	} {
+		if UnsafeArgs(engine, args) == nil {
+			t.Errorf("%s %v was let through", engine, args)
+		}
+	}
+	for _, a := range [][]string{{"--load_format", "pt"}, {"--trust-remote"}, {"--load-form=pt"}} {
+		if UnsafeArgs("vllm", a) == nil {
+			t.Errorf("vllm %v was let through", a)
+		}
+	}
+	// What is not one of them, nor the start of one, still passes.
+	for _, a := range [][]string{{"--max-model-len", "8192"}, {"--tokenizer-mode", "mistral"}, {"--gpu-memory-utilization=0.9"}} {
+		if err := UnsafeArgs("vllm", a); err != nil {
+			t.Errorf("vllm %v was refused: %v", a, err)
+		}
+	}
+	// Short flags are matched as written; llama.cpp takes no abbreviations,
+	// so its --chat-template is not taken for --chat-template-file.
+	for _, a := range [][]string{{"-m", "x"}, {"-hf", "o/r"}, {"-mu", "http://x"}} {
+		if UnsafeArgs("llamacpp", a) == nil {
+			t.Errorf("llamacpp %v was let through", a)
+		}
+	}
+	if err := UnsafeArgs("llamacpp", []string{"--chat-template", "chatml"}); err != nil {
+		t.Errorf("llama.cpp's own --chat-template was refused: %v", err)
+	}
+	// ComfyUI's argparse reads --li as --listen.
+	if UnsafeArgs("comfyui", []string{"--li", "0.0.0.0"}) == nil {
+		t.Error("comfyui --li was let through")
+	}
+}

internal/agent/engines/sglang.go

@@ -102,6 +102,7 @@ func (g *SGLang) serverCommand(s ShardSpec) (Command, error) {
 	if p.ServedModelName != "" {
 		args = append(args, "--served-model-name", p.ServedModelName)
 	}
+	args = append(args, safetensorsOnly(p)...)
 	if s.NNodes > 1 {
 		args = append(args,
 			"--nnodes", strconv.Itoa(s.NNodes),

internal/agent/engines/strata.go

@@ -0,0 +1,184 @@
+package engines
+
+import (
+	"context"
+	"fmt"
+	"net"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strconv"
+	"sync/atomic"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/strata"
+)
+
+// StrataConfig locates a Strata installation.
+type StrataConfig struct {
+	// Dir is Strata's source directory, where setup.py and serve/ are.
+	Dir string `koanf:"dir"`
+	// Python is the interpreter of its environment; Dir's .venv when unset.
+	Python string `koanf:"python"`
+	// Engine is the directory of its prebuilt engine, where the strata
+	// binary and the CUDA libraries it ships are; Dir's engine/ when unset.
+	Engine string `koanf:"engine"`
+}
+
+// Strata runs Strata (github.com/Niko1221/Strata): a mixture-of-experts
+// engine that keeps a model's experts in RAM and the ones used most in VRAM,
+// so a 125B model serves from one 24 GB card. It runs a few model families
+// only, each prepared from its GGUF files by Strata's own setup, so the agent
+// starts llmgw strata-serve, which prepares the model and then becomes
+// Strata's server.
+type Strata struct {
+	cfg       StrataConfig
+	installed atomic.Pointer[string]
+}
+
+// Name identifies the engine in the catalog.
+func (*Strata) Name() string { return "strata" }
+
+// HealthPath answers 200 only once the model is loaded.
+func (*Strata) HealthPath() string { return "/health" }
+
+// MetricsPath is Strata's Prometheus text, in vLLM's names; its /metrics is
+// JSON for its own monitor unless asked for this format.
+func (*Strata) MetricsPath() string { return "/metrics?format=prometheus" }
+
+// ServesHTTP is true: a Strata replica is one process on one node.
+func (*Strata) ServesHTTP(ShardSpec) bool { return true }
+
+// UseInstalled points the engine at the environment the agent installed:
+// its interpreter, with Strata's code beside its bin in strata/ and the
+// engine in strata-engine/.
+func (s *Strata) UseInstalled(python string) { s.installed.Store(&python) }
+
+// Configured reports whether the configuration names a Strata.
+func (s *Strata) Configured() bool { return s.cfg.Dir != "" }
+
+// StrataSourceDir and StrataEngineDir are where an installed environment
+// keeps Strata's code and its prebuilt engine, beside bin/.
+const (
+	StrataSourceDir = "strata"
+	StrataEngineDir = "strata-engine"
+)
+
+func (s *Strata) env() string {
+	if p := s.installed.Load(); p != nil {
+		return filepath.Dir(filepath.Dir(*p))
+	}
+	return ""
+}
+
+func (s *Strata) dir() string {
+	if e := s.env(); e != "" {
+		return filepath.Join(e, StrataSourceDir)
+	}
+	return s.cfg.Dir
+}
+
+func (s *Strata) engineDir() string {
+	if e := s.env(); e != "" {
+		return filepath.Join(e, StrataEngineDir)
+	}
+	if s.cfg.Engine != "" {
+		return s.cfg.Engine
+	}
+	return filepath.Join(s.cfg.Dir, "engine")
+}
+
+func (s *Strata) python() string {
+	if p := s.installed.Load(); p != nil {
+		return *p
+	}
+	if s.cfg.Python != "" {
+		return s.cfg.Python
+	}
+	return filepath.Join(s.cfg.Dir, ".venv", "bin", "python")
+}
+
+var strataVersion = regexp.MustCompile(`PRODUCT_VERSION\s*=\s*"([^"]+)"`)
+
+// Probe reports the Strata a configuration names, by the version its source
+// states: a directory with its setup and server, an interpreter, and the
+// engine binary. Without it a Strata named by hand was never reported, and
+// so never placed on.
+func (s *Strata) Probe(context.Context) (string, error) {
+	dir := s.dir()
+	if dir == "" {
+		return "", fmt.Errorf("no Strata directory is configured")
+	}
+	for _, f := range []string{filepath.Join(dir, "setup.py"), filepath.Join(dir, "serve", "server.py"),
+		s.python(), filepath.Join(s.engineDir(), "strata")} {
+		if _, err := os.Stat(f); err != nil {
+			return "", fmt.Errorf("%s is not a whole Strata: %w", dir, err)
+		}
+	}
+	b, err := os.ReadFile(filepath.Join(dir, "strata_cli", "__init__.py"))
+	if err != nil {
+		return Unknown, nil //nolint:nilerr // whole, of a version it does not say
+	}
+	if m := strataVersion.FindSubmatch(b); m != nil {
+		return string(m[1]), nil
+	}
+	return Unknown, nil
+}
+
+// Validate checks the profile: one card, on one node. Strata splits a model
+// over a node's cards by layer, but that is not driven from here yet.
+func (*Strata) Validate(p Profile, gpus []GPUInfo) error {
+	if p.Nodes > 1 || p.TP > 1 || p.PP > 1 || p.GPUsPerNode > 1 {
+		return fmt.Errorf("profile %q: strata runs on one GPU of one node", p.Name)
+	}
+	// Pascal and Volta run its CUDA 12 build, which the install picks for
+	// them.
+	if err := requireComputeCapability(p, gpus, 6.0, "strata"); err != nil {
+		return err
+	}
+	// A family whose engine needs more: said now, rather than by Strata's
+	// setup after the replica was placed here.
+	if fam, ok := strata.FamilyNamed(strata.ArgValue(p.Args, "--family")); ok && fam.MinCC > 0 {
+		if err := requireComputeCapability(p, gpus, fam.MinCC, "strata's "+fam.Title+" engine"); err != nil {
+			return err
+		}
+	}
+	return validateCommon(p, gpus)
+}
+
+// Commands is llmgw strata-serve, which prepares the model with Strata's
+// setup and then execs Strata's server.
+func (s *Strata) Commands(sh ShardSpec) ([]Command, error) {
+	dir := s.dir()
+	if dir == "" {
+		return nil, fmt.Errorf("strata: no Strata directory is configured (engines.strata.dir)")
+	}
+	bin, err := os.Executable()
+	if err != nil {
+		return nil, err
+	}
+	p := sh.Profile
+	args := []string{"strata-serve",
+		"--listen", net.JoinHostPort(sh.Host(), strconv.Itoa(sh.HTTPPort)),
+		"--model", p.ServedModelName,
+		"--weights", p.ModelPath,
+		// Relative to where engines run — their user's home — so they are
+		// theirs to write. The data is shared by the node's replicas: a
+		// prepared model and its draft layer are gigabytes, the same for
+		// every replica of it.
+		"--state", filepath.Join("strata", sh.ReplicaID),
+		"--data", filepath.Join("strata", "data"),
+		"--python", s.python(),
+		"--strata", dir,
+		"--engine", s.engineDir(),
+		"--",
+	}
+	args = append(args, p.Args...)
+	env := baseEnv(sh)
+	if sh.APIKey != "" {
+		// Strata's server reads its key from here, and strata-serve passes
+		// it on in the environment: not on a command line every user on the
+		// node can read.
+		env["STRATA_API_KEY"] = sh.APIKey
+	}
+	return []Command{{Name: "strata", Path: bin, Args: args, Env: materializeEnv(env)}}, nil
+}

internal/agent/engines/strata_test.go

@@ -0,0 +1,113 @@
+package engines
+
+import (
+	"os"
+	"path/filepath"
+	"slices"
+	"strings"
+	"testing"
+)
+
+func TestStrataStartsStrataServeWithTheKeyInTheEnvironment(t *testing.T) {
+	s := &Strata{}
+	s.UseInstalled("/opt/llmgw/engines/strata/0.2.7-cuda13/bin/python")
+	spec := ShardSpec{
+		ReplicaID: "r1", HTTPPort: 18090, BindAddr: "10.0.0.147", GPUIndices: []int{0},
+		APIKey: "secret-key",
+		Profile: Profile{Name: "p", Engine: "strata", TP: 1, ModelPath: "/var/lib/llmgw/models/q38",
+			ServedModelName: "qwen3.8-flash-next", Args: []string{"--family", "qwen", "--size", "IQ2_XS"}},
+	}
+	cmds, err := s.Commands(spec)
+	if err != nil {
+		t.Fatal(err)
+	}
+	c := cmds[0]
+	args := strings.Join(c.Args, " ")
+	env := "/opt/llmgw/engines/strata/0.2.7-cuda13"
+	for _, want := range []string{
+		"strata-serve", "--listen 10.0.0.147:18090", "--model qwen3.8-flash-next",
+		"--weights /var/lib/llmgw/models/q38", "--state " + filepath.Join("strata", "r1"),
+		"--python " + env + "/bin/python", "--strata " + env + "/strata", "--engine " + env + "/strata-engine",
+		"-- --family qwen --size IQ2_XS",
+	} {
+		if !strings.Contains(args, want) {
+			t.Errorf("args lack %q: %s", want, args)
+		}
+	}
+	if strings.Contains(args, "secret-key") {
+		t.Error("the key is on the command line, which every user on the node can read")
+	}
+	if !slices.Contains(c.Env, "STRATA_API_KEY=secret-key") {
+		t.Error("the key is not in the environment, where Strata's server reads it")
+	}
+	if !slices.Contains(c.Env, "CUDA_VISIBLE_DEVICES=0") {
+		t.Error("the shard is not pinned to its GPU")
+	}
+}
+
+func TestStrataRunsOnOneCard(t *testing.T) {
+	s := &Strata{}
+	p := Profile{Name: "p", Engine: "strata", TP: 1, GPUsPerNode: 2, ModelPath: "/m"}
+	if err := s.Validate(p, nil); err == nil {
+		t.Error("a two-card profile was accepted")
+	}
+	p.GPUsPerNode = 1
+	if err := s.Validate(p, []GPUInfo{{Name: "Quadro RTX 6000", ComputeCapability: "7.5"}}); err != nil {
+		t.Errorf("one Turing card was refused: %v", err)
+	}
+}
+
+func TestStrataSetupFlagsThatMoveFilesAreRefused(t *testing.T) {
+	for _, bad := range []string{"--gguf-dir", "--gguf", "--data-dir=/etc", "--model", "--mod", "--feed", "--host", "--gpus"} {
+		if err := UnsafeArgs("strata", []string{bad, "x"}); err == nil {
+			t.Errorf("%s was accepted", bad)
+		}
+	}
+	if err := UnsafeArgs("strata", []string{"--family", "qwen", "--size", "IQ2_XS", "--context", "32768",
+		"--parallel", "2", "--vram-reserve-mib", "2048", "--vision", "no"}); err != nil {
+		t.Errorf("an ordinary profile was refused: %v", err)
+	}
+}
+
+func TestAStrataNamedByHandIsReportedWithItsVersion(t *testing.T) {
+	dir := t.TempDir()
+	for _, f := range []string{"setup.py", "serve/server.py", ".venv/bin/python", "engine/strata"} {
+		p := filepath.Join(dir, f)
+		if err := os.MkdirAll(filepath.Dir(p), 0o755); err != nil {
+			t.Fatal(err)
+		}
+		if err := os.WriteFile(p, nil, 0o755); err != nil {
+			t.Fatal(err)
+		}
+	}
+	s := &Strata{cfg: StrataConfig{Dir: dir}}
+	if v, err := s.Probe(t.Context()); err != nil || v != Unknown {
+		t.Fatalf("%q %v", v, err)
+	}
+	_ = os.MkdirAll(filepath.Join(dir, "strata_cli"), 0o755)
+	_ = os.WriteFile(filepath.Join(dir, "strata_cli", "__init__.py"), []byte("PRODUCT_VERSION = \"0.2.7\"\n"), 0o644)
+	if v, _ := s.Probe(t.Context()); v != "0.2.7" {
+		t.Errorf("version %q", v)
+	}
+	r := NewRegistry(Config{Strata: StrataConfig{Dir: dir}})
+	if got := r.ProbeConfigured(t.Context()); got["strata"] != "0.2.7" {
+		t.Errorf("not reported: %v", got)
+	}
+	_ = os.Remove(filepath.Join(dir, "engine", "strata"))
+	if _, err := s.Probe(t.Context()); err == nil {
+		t.Error("reported without its engine")
+	}
+}
+
+func TestStrataRefusesDeepSeekOnACardItsEngineWasNotBuiltFor(t *testing.T) {
+	s := &Strata{}
+	p := Profile{Name: "p", Engine: "strata", ModelPath: "/m", TP: 1, GPUsPerNode: 1, Args: []string{"--family", "deepseek41", "--size", "IST-DAS-1.75bpw"}}
+	gpus := []GPUInfo{{Index: 0, Name: "Quadro RTX 6000", ComputeCapability: "7.5"}}
+	if err := s.Validate(p, gpus); err == nil || !strings.Contains(err.Error(), "8.6") {
+		t.Errorf("DeepSeek on 7.5: %v", err)
+	}
+	p.Args = []string{"--family", "qwen", "--size", "IQ2_XS"}
+	if err := s.Validate(p, gpus); err != nil {
+		t.Errorf("Qwen on 7.5: %v", err)
+	}
+}

internal/agent/engines/unsafeenv.go

@@ -2,6 +2,7 @@ package engines
 
 import (
 	"fmt"
+	"os"
 	"strings"
 )
 
@@ -24,7 +25,7 @@ var envFamilies = []string{
 	"CUDA_", "NCCL_", "NVIDIA_", "NVTE_", "GLOO_", "UCX_", "TORCH_", "PYTORCH_",
 	"VLLM_", "SGLANG_", "GGML_", "LLAMA_", "OMP_", "MKL_", "OPENBLAS_", "KMP_",
 	"HF_", "TRANSFORMERS_", "TOKENIZERS_", "TRITON_", "FLASHINFER_",
-	"TRTLLM_", "TLLM_",
+	"TRTLLM_", "TLLM_", "FREETOKEN_",
 }
 
 // envExact are single variables that are tuning too.
@@ -32,6 +33,14 @@ var envExact = map[string]bool{
 	"PYTHONUNBUFFERED": true, "PYTHONHASHSEED": true, "TZ": true,
 }
 
+// envNeverExact are variables in an allowed family that no profile may set,
+// whose names the patterns below do not catch: what the agent sets itself,
+// and FreeToken's way of loading a projector file.
+var envNeverExact = map[string]string{
+	"FREETOKEN_API_KEY": "is set by the agent", "FREETOKEN_HOST": "is set by the agent",
+	"FREETOKEN_MMPROJ": "loads a file",
+}
+
 // envNever are names in an allowed family that load code or move where the
 // engine reads and writes.
 var envNever = []string{
@@ -58,6 +67,9 @@ func UnsafeEnv(key string) (bool, string) {
 		return true, "only the variables engines are tuned with are allowed (" +
 			strings.Join(envFamilies, "*, ") + "*)"
 	}
+	if why, ok := envNeverExact[k]; ok {
+		return true, "it " + why
+	}
 	for _, n := range envNever {
 		if strings.Contains(k, n) && !envExact[k] {
 			return true, "it names a path, a plugin, a library or an address, which is a way to make the engine " +
@@ -101,6 +113,13 @@ var flagsNever = map[string]map[string]string{
 		"--hf-token": "is a credential", "--alias": "is set by the agent",
 		"--model-draft": "loads a file", "-md": "loads a file",
 		"--prompt-cache": "writes a file", "--logdir": "writes files",
+		// The short spellings of the same.
+		"-mm": "loads a file; the projector is fetched with the model", "--mmproj-url": "downloads a file",
+		"-mmu": "downloads a file", "-hfr": "downloads a model", "-hff": "downloads a model",
+		"--hf-repo-draft": "downloads a model", "-hfd": "downloads a model", "-hfrd": "downloads a model",
+		"--hf-repo-v": "downloads a model", "-hfv": "downloads a model", "--hf-file-v": "downloads a model",
+		"-hffv": "downloads a model", "-hft": "is a credential", "-jf": "reads a file", "-a": "is set by the agent",
+		"--model-vocoder": "loads a file", "-mv": "loads a file",
 	},
 	"vllm": {
 		"--download-dir": "writes files", "--tokenizer": "reads another repository or path",
@@ -118,6 +137,7 @@ var flagsNever = map[string]map[string]string{
 		"--file-storage-path": "writes files", "--log-requests": "logs prompts",
 		"--model-path": "is the model's source, set by the agent", "--served-model-name": "is set by the agent",
 		"--lora-paths": "loads files", "--download-dir": "writes files",
+		"--load-format": "changes how weights are loaded, some ways running code",
 	},
 	// ComfyUI's main.py, behind the gate that sets where it listens and
 	// where its files are.
@@ -132,6 +152,32 @@ var flagsNever = map[string]map[string]string{
 		"--tls-keyfile":        "reads a file", "--tls-certfile": "reads a file",
 		"--comfy-api-base": "sends work to another service",
 	},
+	// Strata's setup.py, which a profile's arguments go to: llmgw
+	// strata-serve sets where the model is, where its files go and where
+	// the server listens, and the engine is the one the agent installed.
+	// Its own --model, the size, is spelt --size in a profile, since
+	// --model is the source everywhere else.
+	"strata": {
+		"--gguf-dir": "is the model's source, set by the agent", "--models-dir": "is set by the agent",
+		"--data-dir": "is set by the agent", "--move-data": "moves files", "--set-default-data-dir": "writes settings",
+		"--feed": "downloads an engine", "--prebuilt": "loads an engine", "--build": "compiles an engine",
+		"--update": "downloads an engine", "--rollback-engine": "changes the engine",
+		"--dsv41-package": "loads a file", "--inspect": "reads another repository or path",
+		"--source": "downloads from elsewhere", "--setup": "is not for a server",
+		"--browser": "opens a browser", "--gpu": "is set by the agent", "--gpus": "is set by the agent",
+		"--gpu-roles": "is set by the agent", "--backend": "is set by the agent",
+		"--experimental-speed-projection": "changes what the model refuses",
+	},
+	// ft serve: where it listens, its key, its model and its card are the
+	// agent's, and what reads files or fetches is refused.
+	"freetoken": {
+		"--gpu": "is set by the agent", "--served-model-name": "is set by the agent",
+		"--mmproj": "loads a file", "--mm-encoder-weights": "loads a file",
+		"--allowed-local-media-path": "serves local files to requests",
+		"--allowed-media-domains":    "lets requests have the engine fetch from those sites",
+		"--moe-cache-split":          "reads a profile file",
+		"--hf-overrides":             "rewrites the checkpoint's configuration, which can name code to load",
+	},
 	// trtllm-serve spells its flags with underscores, so the shared list's
 	// hyphenated names do not reach them.
 	"trtllm": {
@@ -163,14 +209,43 @@ var flagsNever = map[string]map[string]string{
 // unsafeFlag is why an engine may not be given a flag, or "".
 func unsafeFlag(engine, flag, _ string) string {
 	name, _, _ := strings.Cut(flag, "=")
+	if !strings.HasPrefix(name, "-") {
+		return ""
+	}
+	// Written as the engine would read it: vLLM takes --load_format for
+	// --load-format, TensorRT-LLM the other way round. Short flags (-m,
+	// -hf) are matched as written.
+	norm := func(f string) string {
+		if strings.HasPrefix(f, "--") {
+			return "--" + strings.ReplaceAll(f[2:], "_", "-")
+		}
+		return f
+	}
+	given := norm(name)
+	// argparse also takes any unambiguous beginning of a long flag for the
+	// flag — --trust-remote is --trust-remote-code, --li is ComfyUI's
+	// --listen — so for the engines that parse with it, a beginning of a
+	// refused flag is refused. llama.cpp and TensorRT-LLM take no such
+	// thing, and --chat-template there is a flag of its own.
+	abbreviates := argparseEngines[engine] && strings.HasPrefix(given, "--") && len(given) >= 4
 	for _, set := range []string{"", engine} {
-		if why, ok := flagsNever[set][name]; ok && why != "" {
-			return why
+		for never, why := range flagsNever[set] {
+			if why == "" {
+				continue
+			}
+			n := norm(never)
+			if n == given || (abbreviates && strings.HasPrefix(n, "--") && strings.HasPrefix(n, given)) {
+				return why
+			}
 		}
 	}
 	return ""
 }
 
+// argparseEngines read their flags with Python's argparse, which takes a
+// long flag's beginning for the flag.
+var argparseEngines = map[string]bool{"vllm": true, "sglang": true, "comfyui": true, "strata": true, "freetoken": true}
+
 // UnsafeArgs refuses the first of a profile's arguments an engine may not be
 // given, saying why.
 func UnsafeArgs(engine string, args []string) error {
@@ -185,3 +260,29 @@ func UnsafeArgs(engine string, args []string) error {
 	}
 	return nil
 }
+
+// safetensorsOnly has vLLM or SGLang read only safetensors from a model it
+// downloads itself, by its Hugging Face id: left to choose, either takes
+// the repository's pickle files (pytorch_model.bin) when it has no
+// safetensors, and loading those runs whatever they hold. A path was
+// fetched by the agent, which checked it, or put there by the operator.
+func safetensorsOnly(p Profile) []string {
+	if p.AllowPickle || !hubID(p.ModelPath) {
+		return nil
+	}
+	return []string{"--load-format", "safetensors"}
+}
+
+// hubID reports whether a model path is a Hugging Face id, org/name, rather
+// than a file or directory.
+func hubID(s string) bool {
+	if s == "" || strings.HasPrefix(s, "/") || strings.HasPrefix(s, ".") || strings.Contains(s, "://") {
+		return false
+	}
+	// org/name:Q4_K_M is a GGUF in a repository, which vLLM reads as one.
+	if strings.Count(s, "/") != 1 || strings.Contains(s, ":") {
+		return false
+	}
+	_, err := os.Stat(s)
+	return err != nil
+}

internal/agent/engines/vllm.go

@@ -119,6 +119,7 @@ func (v *VLLM) serverCommand(s ShardSpec) (Command, error) {
 	if p.ServedModelName != "" {
 		args = append(args, "--served-model-name", p.ServedModelName)
 	}
+	args = append(args, safetensorsOnly(p)...)
 
 	env := baseEnv(s)
 

internal/agent/firewall.go

@@ -10,8 +10,20 @@ import (
 
 	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
 	"gitea.mynodes.xyz/ilker/llmgw/internal/firewall"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
 )
 
+// slurmPorts are Slurm's on a node that runs any of it: slurmctld's and
+// slurmd's, and srun's range for the steps it starts.
+var slurmPorts = []firewall.Port{
+	{From: slurm.CtldPort, To: slurm.DbdPort, What: "slurm: slurmctld, slurmd and slurmdbd"},
+	{From: slurm.SrunPortsFrom, To: slurm.SrunPortsTo, What: "slurm: srun, for the steps it starts"},
+}
+
+// nfsPort is NFSv4's, on the machine that exports the Slurm cluster's
+// shared files.
+var nfsPort = firewall.Port{From: slurm.NFSPort, To: slurm.NFSPort, What: "nfs: the Slurm cluster's shared files"}
+
 // applyFirewall keeps llmgw's firewall on this node as the cluster's
 // settings say: the engines' ports and the node exporter's for the networks
 // the node reaches its control planes on, and those named, or no table at
@@ -19,6 +31,7 @@ import (
 func (a *Agent) applyFirewall(m *llmgwv1.Firewall) {
 	a.fwMu.Lock()
 	defer a.fwMu.Unlock()
+	a.fwLast = m
 	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
 	defer cancel()
 	r := a.fwRunner
@@ -55,6 +68,14 @@ func (a *Agent) applyFirewall(m *llmgwv1.Firewall) {
 	if from > 0 && to >= from {
 		ports = append(ports, firewall.Engines(from, to))
 	}
+	// Once there, they stay: the ruleset keeps every port it held, so a
+	// node that leaves Slurm keeps them closed to everyone but the cluster.
+	if a.slurm != nil && a.slurm.HasPart() {
+		ports = append(ports, slurmPorts...)
+	}
+	if a.slurm != nil && a.slurm.Serves() {
+		ports = append(ports, nfsPort)
+	}
 	changed, err := firewall.Install(ctx, r, firewall.Spec{Ports: ports, Allow: nets})
 	switch {
 	case err != nil:
@@ -64,6 +85,23 @@ func (a *Agent) applyFirewall(m *llmgwv1.Firewall) {
 	}
 }
 
+// slurmMessage is the leader's word on this node's part in Slurm. A node
+// that gets one, or starts or stops exporting the shared files, has the
+// firewall applied again, with Slurm's ports and NFS's.
+func (a *Agent) slurmMessage(m *llmgwv1.Slurm) {
+	had, served := a.slurm.HasPart(), a.slurm.Serves()
+	a.slurm.Apply(m)
+	if (had || !a.slurm.HasPart()) && served == a.slurm.Serves() {
+		return
+	}
+	a.fwMu.Lock()
+	last := a.fwLast
+	a.fwMu.Unlock()
+	if last != nil && last.GetEnabled() {
+		go a.applyFirewall(last)
+	}
+}
+
 // controlPlaneIPs are the addresses of the control planes an agent dials.
 func controlPlaneIPs(ctx context.Context, servers []string) []string {
 	var out []string

internal/agent/firewall_test.go

@@ -6,8 +6,10 @@ import (
 	"log/slog"
 	"strings"
 	"testing"
+	"time"
 
 	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/slurmnode"
 )
 
 // fakeShell answers the firewall package's scripts and records what was
@@ -55,3 +57,28 @@ func TestANodeKeepsOrDropsItsFirewallAsTheClusterSays(t *testing.T) {
 		t.Error("turned off, the table stayed")
 	}
 }
+
+func TestANodeGivenAPartInSlurmLetsSlurmInFromTheClusterOnly(t *testing.T) {
+	sh := &fakeShell{}
+	a := &Agent{log: slog.New(slog.DiscardHandler), fwRunner: sh, serverList: []string{"https://10.0.0.140:7443"},
+		slurm: slurmnode.New(nil, nil, "", slog.New(slog.DiscardHandler))}
+	a.applyFirewall(&llmgwv1.Firewall{Enabled: true, EnginePortsFrom: 20000, EnginePortsTo: 20999})
+	if strings.Contains(sh.installed, "6817") {
+		t.Fatalf("Slurm's ports before the node has a part in it:\n%s", sh.installed)
+	}
+
+	a.slurmMessage(&llmgwv1.Slurm{Name: "node05", Node: true, ConfServer: "10.0.0.148:6817"})
+	deadline := time.Now().Add(5 * time.Second)
+	for {
+		a.fwMu.Lock()
+		got := sh.installed
+		a.fwMu.Unlock()
+		if strings.Contains(got, "6817-6819") && strings.Contains(got, "60001-63000") {
+			break
+		}
+		if time.Now().After(deadline) {
+			t.Fatalf("given a part in Slurm, the rules still lack its ports:\n%s", got)
+		}
+		time.Sleep(10 * time.Millisecond)
+	}
+}

internal/agent/gpucheck.go

@@ -20,6 +20,7 @@ var (
 	gpuProcesses = discovery.GPUProcesses
 	gpuMemory    = discovery.GPUMemoryByProcess
 	processGroup = pgidOf
+	parentOf     = ppidOf
 	// gpuCheckWait is how long a shard has, once it answers, to show up on
 	// its GPUs: the weights are on them by then, but nvidia-smi can lag.
 	gpuCheckWait = 15 * time.Second
@@ -116,19 +117,44 @@ func (a *Agent) foreignGPUMemory(ctx context.Context, shards []int) (foreign, us
 	return foreign, used
 }
 
-// holds reports whether any of pids is the shard's process or in its group:
-// an engine that computes in a child process is still the shard.
+// holds reports whether any of pids is the shard's process, in its group,
+// or a process it started: an engine that computes in a child process is
+// still the shard, and Strata's server starts its engine in a group of the
+// engine's own.
 func holds(pids []int, pid, group int) bool {
 	for _, p := range pids {
-		if p == pid || (group > 0 && processGroup(p) == group) {
+		if p == pid || (group > 0 && processGroup(p) == group) || descends(p, pid) {
 			return true
 		}
 	}
 	return false
 }
 
+// descends reports whether process p was started, at any remove, by pid.
+func descends(p, pid int) bool {
+	// Bounded: a chain this deep is no engine's, and a /proc read racing an
+	// exit must not loop.
+	for range 16 {
+		p = parentOf(p)
+		switch {
+		case p == pid:
+			return true
+		case p <= 1:
+			return false
+		}
+	}
+	return false
+}
+
 // pgidOf reads a process's group from /proc; zero when it cannot.
-func pgidOf(pid int) int {
+func pgidOf(pid int) int { return statField(pid, 2) }
+
+// ppidOf reads a process's parent from /proc; zero when it cannot.
+func ppidOf(pid int) int { return statField(pid, 1) }
+
+// statField is one of the numbers after the command name in /proc's stat
+// line: 1 is the parent, 2 the group.
+func statField(pid, field int) int {
 	b, err := os.ReadFile("/proc/" + strconv.Itoa(pid) + "/stat")
 	if err != nil {
 		return 0
@@ -141,10 +167,10 @@ func pgidOf(pid int) int {
 		return 0
 	}
 	f := strings.Fields(s[i+1:])
-	if len(f) < 3 {
+	if len(f) <= field {
 		return 0
 	}
-	n, _ := strconv.Atoi(f[2])
+	n, _ := strconv.Atoi(f[field])
 	return n
 }
 

internal/agent/gpucheck_test.go

@@ -123,6 +123,30 @@ func TestForeignGPUMemoryLeavesOutTheAgentsOwnShards(t *testing.T) {
 	}
 }
 
+func TestAnEngineStartedInAGroupOfItsOwnIsStillTheShards(t *testing.T) {
+	// cuda01: Strata's server (259709) starts its engine (261072) in a
+	// group of the engine's own, and the engine holds the card. It read as
+	// a shard on the CPU, and as 22 GB of somebody else's memory.
+	oldMem, oldGroup, oldParent := gpuMemory, processGroup, parentOf
+	t.Cleanup(func() { gpuMemory, processGroup, parentOf = oldMem, oldGroup, oldParent })
+	processGroup = func(pid int) int { return pid }
+	parentOf = func(pid int) int {
+		return map[int]int{261072: 259709, 259709: 258734, 258734: 1, 999: 1}[pid]
+	}
+	gpuMemory = func(context.Context, string) ([]discovery.GPUMemory, error) {
+		return []discovery.GPUMemory{{Index: 0, UsedMB: 23000,
+			Procs: []discovery.GPUProcess{{PID: 261072, UsedMB: 22226}, {PID: 999, UsedMB: 700}}}}, nil
+	}
+	a := &Agent{log: slog.Default()}
+	if c, ok := a.gpuCompute(context.Background(), []int{0}, 259709); !ok || c.Backend != "CUDA" ||
+		len(c.Devices) != 1 || c.Devices[0] != "GPU 0: 21.7 GB" {
+		t.Errorf("the engine's child on GPU 0: %+v %v", c, ok)
+	}
+	if foreign, _ := a.foreignGPUMemory(context.Background(), []int{259709}); !slices.Equal(foreign, []int32{774}) {
+		t.Errorf("foreign %v; want only what the other process and nobody holds, [774]", foreign)
+	}
+}
+
 func TestLlamaCppsOwnWordOnTheCPUIsEnoughToRefuse(t *testing.T) {
 	eng := fallbackEngine{}
 	// Without nvidia-smi to ask, the engine's own output says it.

internal/agent/main_test.go

@@ -0,0 +1,17 @@
+package agent
+
+import (
+	"os"
+	"testing"
+	"time"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
+)
+
+func TestMain(m *testing.M) {
+	// Half a gigabyte read per test binary, and contending with the tests
+	// for the CPUs: these never look at the figure.
+	measureBandwidth = func() discovery.Bandwidth { return discovery.Bandwidth{} }
+	bandwidthDelay = func() time.Duration { return 0 }
+	os.Exit(m.Run())
+}

internal/agent/provision.go

@@ -164,6 +164,13 @@ func (a *Agent) provisionLoop(ctx context.Context) {
 
 // provisionStatus records a change in one engine's installation.
 func (a *Agent) provisionStatus(st provision.Status) {
+	if st == (provision.Status{}) {
+		// A package install that was no engine's has ended.
+		a.mu.Lock()
+		delete(a.provStatus, "")
+		a.mu.Unlock()
+		return
+	}
 	a.mu.Lock()
 	prev, had := a.provStatus[st.Engine]
 	a.provStatus[st.Engine] = st

internal/agent/provision/freetoken_test.go

@@ -0,0 +1,53 @@
+package provision
+
+import (
+	"errors"
+	"slices"
+	"strings"
+	"testing"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
+)
+
+func TestFreeTokenIsInstalledOnlyWhereItsKernelsRun(t *testing.T) {
+	old := freetokenToolkit
+	toolkit := "/usr/local/cuda-13.4"
+	freetokenToolkit = func() string { return toolkit }
+	t.Cleanup(func() { freetokenToolkit = old })
+
+	ampere := discovery.Host{Arch: "amd64", NVIDIADriver: "580.95.05", MinComputeCapability: 8.6}
+	got, err := freetokenCandidates("0.2.1", "", ampere)
+	if err != nil {
+		t.Fatal(err)
+	}
+	b := got[0]
+	if b.Binary != "ft" || !slices.Equal(b.Requirements, []string{"freetoken[accel]==0.2.1"}) || !b.GPUCheck || b.Python == "" {
+		t.Errorf("build %+v", b)
+	}
+
+	// Not meant for these, and not reported as a fault: no GPU, a Turing
+	// card (cuda01, where 0.2.1 had no kernel image for it), an Arm node.
+	for name, h := range map[string]discovery.Host{
+		"no gpu":  {Arch: "amd64"},
+		"turing":  {Arch: "amd64", NVIDIADriver: "615.71.09", MinComputeCapability: 7.5},
+		"aarch64": {Arch: "arm64", NVIDIADriver: "580.95.05", MinComputeCapability: 9.0},
+	} {
+		_, err := freetokenCandidates("0.2.1", "", h)
+		var nh NotHere
+		if !errors.As(err, &nh) {
+			t.Errorf("%s: %v, want NotHere", name, err)
+		}
+	}
+
+	// Meant for these, and said why not.
+	if _, err := freetokenCandidates("0.2.1", "", discovery.Host{Arch: "amd64", NVIDIADriver: "570.1", MinComputeCapability: 9.0}); err == nil ||
+		!strings.Contains(err.Error(), "driver 580") {
+		t.Errorf("old driver: %v", err)
+	}
+	toolkit = ""
+	_, err = freetokenCandidates("0.2.1", "", ampere)
+	var nh NotHere
+	if err == nil || errors.As(err, &nh) || !strings.Contains(err.Error(), "CUDA 13 toolkit") {
+		t.Errorf("no toolkit: %v", err)
+	}
+}

internal/agent/provision/install.go

@@ -174,7 +174,7 @@ func (in *Installer) one(ctx context.Context, w Want) Status {
 		}
 		return st
 	}
-	if w.Offline != nil && recipe.Repo == "" {
+	if w.Offline != nil && (recipe.Repo == "" || builds[0].Python != "") {
 		// Said, not attempted: uv would spend minutes failing to reach PyPI,
 		// and the failure would read like a network fault rather than the
 		// design of an offline site.
@@ -483,7 +483,8 @@ func (in *Installer) packages(ctx context.Context, byFamily map[string][]string,
 		return fmt.Errorf("%s is missing and no package manager was found to install it",
 			strings.Join(missing, ", "))
 	}
-	in.progress(st, "installing %s with %s", strings.Join(missing, ", "), tool)
+	// The state already says installing.
+	in.progress(st, "%s, with %s", strings.Join(missing, ", "), tool)
 	for _, c := range cmds {
 		if err := in.runWaitingForLock(ctx, pm.env, c, st); err != nil {
 			return err
@@ -497,7 +498,13 @@ func (in *Installer) packages(ctx context.Context, byFamily map[string][]string,
 // byFamily is keyed by discovery.Host.Family: "deb" or "rpm".
 func (in *Installer) InstallPackages(ctx context.Context, byFamily map[string][]string) error {
 	var st Status
-	return in.packages(ctx, byFamily, &st)
+	err := in.packages(ctx, byFamily, &st)
+	if st.State == Installing {
+		// Its progress was no engine's: an empty status takes it off the
+		// node's page again, done or failed (the caller says which).
+		in.report(Status{})
+	}
+	return err
 }
 
 // lockWait bounds how long a package command waits for another to finish.

internal/agent/provision/packages_status_test.go

@@ -0,0 +1,36 @@
+package provision
+
+import (
+	"context"
+	"testing"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
+)
+
+// A package install that is no engine's says it is done, so its progress
+// does not stay on the node's page (found on the PoC installing Slurm).
+func TestInstallPackagesClearsItsProgress(t *testing.T) {
+	var got []Status
+	in := &Installer{Host: discovery.Host{OSID: "ubuntu"}, OnStatus: func(s Status) { got = append(got, s) },
+		Run: func(_ context.Context, _ []string, name string, _ ...string) ([]byte, error) {
+			if name == "sh" { // the query: not installed
+				return nil, context.Canceled
+			}
+			return nil, nil
+		}}
+	if err := in.InstallPackages(context.Background(), map[string][]string{"deb": {"munge"}}); err != nil {
+		t.Fatal(err)
+	}
+	if len(got) < 2 || got[0].State != Installing || got[len(got)-1] != (Status{}) {
+		t.Fatalf("statuses %+v", got)
+	}
+
+	got = nil
+	in.Run = func(context.Context, []string, string, ...string) ([]byte, error) { return nil, nil }
+	if err := in.InstallPackages(context.Background(), map[string][]string{"deb": {"munge"}}); err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 0 {
+		t.Fatalf("installed already, yet said %+v", got)
+	}
+}

internal/agent/provision/python.go

@@ -109,6 +109,11 @@ func (in *Installer) installPython(ctx context.Context, r Recipe, w Want, b Buil
 		}
 		reqs = append(reqs, more...)
 	}
+	if len(b.Assets) > 0 {
+		if err := in.unpackAssets(ctx, r, w, b, tmp, st); err != nil {
+			return err
+		}
+	}
 	if b.Wheel != nil {
 		// The engine from its GitHub release, pinned to the digest GitHub
 		// publishes for the file: the same trust as an archive, and uv
@@ -372,6 +377,48 @@ func (in *Installer) unpackSource(ctx context.Context, w Want, b Build, env stri
 	return requirementsIn(string(f), b.IndexPackages)
 }
 
+// unpackAssets fetches a Python engine's release binaries, each checked
+// against the digest GitHub publishes for it, and unpacks them into the
+// environment's AssetDir.
+func (in *Installer) unpackAssets(ctx context.Context, r Recipe, w Want, b Build, env string, st *Status) error {
+	tag := b.AssetTag
+	if tag == "" {
+		tag = w.Version
+	}
+	in.progress(st, "asking GitHub for the checksums of %s %s", r.Repo, tag)
+	sums, err := in.checksums(ctx, r.Repo, tag)
+	if err != nil {
+		return err
+	}
+	dst := filepath.Join(env, b.AssetDir)
+	if err := os.MkdirAll(dst, 0o755); err != nil {
+		return err
+	}
+	for _, asset := range b.Assets {
+		want, ok := sums[asset]
+		if !ok {
+			// Refused rather than installed unchecked: this is a binary that
+			// will run on every node.
+			return fmt.Errorf("%s %s publishes no checksum for %s, so it cannot be verified", r.Repo, tag, asset)
+		}
+		url := strings.TrimSuffix(in.downloadBase(), "/") + "/" + r.Repo + "/releases/download/" + tag + "/" + asset
+		archive := filepath.Join(env+".source", asset)
+		if err := os.MkdirAll(filepath.Dir(archive), 0o755); err != nil {
+			return err
+		}
+		if err := in.download(ctx, url, asset, archive, want, st); err != nil {
+			return err
+		}
+		in.progress(st, "unpacking %s", asset)
+		err := untarGz(archive, dst)
+		_ = os.RemoveAll(filepath.Dir(archive))
+		if err != nil {
+			return fmt.Errorf("unpacking %s: %w", asset, err)
+		}
+	}
+	return nil
+}
+
 // requirementsIn reads a requirements file's lines. Options (-r, -e, --index-url
 // and the like) are refused: they would fetch from somewhere other than PyPI,
 // or install what the lock never saw.

internal/agent/provision/recipe.go

@@ -76,6 +76,11 @@ type Build struct {
 	// was published, which PyPI cannot be asked for.
 	Source *Source
 	AsOf   time.Time
+	// AssetTag and AssetDir, for a Python engine that also ships a binary
+	// in its releases, are the release tag Assets are taken from, checked
+	// against the digests GitHub publishes, and the directory in the
+	// environment they are unpacked into.
+	AssetTag, AssetDir string
 	// FreeGB is the disk space the installation needs before it starts: a
 	// Python engine is several gigabytes, and one that fills the disk halfway
 	// leaves the node worse off than one that never began.
@@ -121,11 +126,13 @@ type Recipe struct {
 // from PyPI by uv, which is itself installed like llama.cpp: a GitHub release
 // checked against its published checksum.
 var Recipes = map[string]Recipe{
-	"llamacpp": {Repo: "ggml-org/llama.cpp", Candidates: llamaCppCandidates},
-	"vllm":     {PyPI: "vllm", Candidates: vllmCandidates},
-	"sglang":   {PyPI: "sglang", Candidates: sglangCandidates},
-	"comfyui":  {Candidates: comfyuiCandidates},
-	"uv":       {Repo: "astral-sh/uv", Candidates: uvCandidates},
+	"llamacpp":  {Repo: "ggml-org/llama.cpp", Candidates: llamaCppCandidates},
+	"vllm":      {PyPI: "vllm", Candidates: vllmCandidates},
+	"sglang":    {PyPI: "sglang", Candidates: sglangCandidates},
+	"comfyui":   {Candidates: comfyuiCandidates},
+	"strata":    {Repo: StrataRepo, Candidates: strataCandidates},
+	"freetoken": {PyPI: "freetoken", Candidates: freetokenCandidates},
+	"uv":        {Repo: "astral-sh/uv", Candidates: uvCandidates},
 }
 
 // Installable are the engines this agent can install, as it tells the
@@ -172,6 +179,9 @@ func Builds(engine, version, arch string) ([]Build, error) {
 			continue
 		}
 		for _, b := range bs {
+			if b.Python != "" {
+				return nil, fmt.Errorf("%s is a Python environment from PyPI, which an offline bundle does not carry", engine)
+			}
 			if !seen[b.Variant] {
 				seen[b.Variant] = true
 				out = append(out, b)
@@ -672,3 +682,133 @@ print("comfyui version", "` + version + `", "on torch", torch.__version__)`},
 		{cu126, gpuNeeds{minDriverForCUDA12, 5.0}, "ComfyUI " + version + " on CUDA 12.6's torch"},
 	}, h)
 }
+
+// --- Strata ---
+
+// StrataRepo is where Strata is published: its source, and in each release
+// the prebuilt engine for each CUDA.
+const StrataRepo = "Niko1221/Strata"
+
+// strataReleases are the Strata versions this agent installs: the source
+// archive's digest, taken when it was added here and confirmed from a second
+// machine, when the version was published, and the engine its release
+// carries, whose files GitHub publishes digests for.
+var strataReleases = map[string]struct {
+	sha256    string
+	published string
+	engine    string
+}{
+	"0.2.7": {"41ae861478f5e5d9506864f97dfb5dd7cfe6b83193300e2b0d4f04c0380e3af9", "2026-10-27T14:54:42Z", "0.1.52"},
+}
+
+// strataCandidates are the Strata builds for a machine: its source, in a
+// Python environment, with the prebuilt engine for the node's driver and
+// cards — built for CUDA 13, Turing to Blackwell; or for CUDA 12.9, Pascal
+// to Ada, which is what a V100, or a driver older than 580, gets.
+func strataCandidates(version, variant string, h discovery.Host) ([]Build, error) {
+	if _, err := gpuArch("Strata", h); err != nil {
+		return nil, err
+	}
+	if h.Arch != "amd64" {
+		return nil, NotHere{"Strata publishes its Linux CUDA engine for x86-64 only"}
+	}
+	rel, ok := strataReleases[version]
+	if !ok {
+		known := make([]string, 0, len(strataReleases))
+		for v := range strataReleases {
+			known = append(known, v)
+		}
+		slices.Sort(known)
+		return nil, fmt.Errorf("this agent knows the source digest of Strata %s, not of %s; "+
+			"install that one by hand and name it in agent.yaml (engines.strata.dir)",
+			strings.Join(known, ", "), version)
+	}
+	published, err := time.Parse(time.RFC3339, rel.published)
+	if err != nil {
+		return nil, err
+	}
+	build := func(cuda string) Build {
+		return Build{
+			Variant: cuda,
+			Python:  pythonForEngines,
+			Source: &Source{
+				URL:    "/" + StrataRepo + "/archive/refs/tags/v" + version + ".tar.gz",
+				SHA256: rel.sha256, Top: "Strata-" + version, Dir: "strata",
+				Requirements: "requirements.txt",
+			},
+			AsOf:     published.Add(24 * time.Hour),
+			Assets:   []string{"strata-engine-" + rel.engine + "-linux-x86_64-cuda-" + cuda + ".tar.gz"},
+			AssetTag: "v" + version,
+			AssetDir: "strata-engine",
+			// Strata is started as a script, so what the agent is given is
+			// the environment's interpreter; the proof is importing what its
+			// server needs, and finding its engine beside it.
+			Binary: "python",
+			VersionArgs: []string{"-c", `import os, sys, numpy, jinja2, regex
+e = os.path.dirname(os.path.dirname(os.path.abspath(sys.executable)))
+for f in ("strata/setup.py", "strata/serve/server.py", "strata-engine/strata"):
+    assert os.path.isfile(os.path.join(e, f)), f + " is missing"
+print("strata version", "` + version + `", "engine", "` + rel.engine + `")`},
+			FreeGB: 5,
+		}
+	}
+	return pick("Strata "+version, variant, []candidate{
+		{build("cuda13"), gpuNeeds{minDriverForCUDA13, 7.5}, "Strata " + version + " built for CUDA 13"},
+		// 12.9 rather than 12: its machine code is for 6.0 to 8.9 and its
+		// PTX was compiled by 12.9, so the driver must be 575.
+		{build("cuda12"), gpuNeeds{575, 6.0}, "Strata " + version + " built for CUDA 12.9"},
+	}, h)
+}
+
+// --- FreeToken ---
+
+// FreeTokenRepo is where FreeToken is developed; its builds are on PyPI.
+const FreeTokenRepo = "FlashML-org/FreeToken"
+
+// freetokenMinCompute is the oldest GPU FreeToken runs on. It says RTX 30
+// and newer, and its kernels agree: on a Quadro RTX 6000 (7.5) version 0.2.1
+// loaded gpt-oss-20b and failed its first prefill with no kernel image for
+// the card, and in fp16 with fused experts its first request hit an illegal
+// memory access. Its FP8 kernels need 8.9 and its NVFP4 ones 12.0, which
+// the model's quantisation decides.
+const freetokenMinCompute = 8.0
+
+// freetokenToolkit finds the CUDA toolkit FreeToken compiles its kernels
+// with; a variable for tests.
+var freetokenToolkit = func() string { return discovery.CUDAToolkit(13) }
+
+// freetokenCandidates are the FreeToken builds for a machine: its PyPI
+// wheels with the kernel extras (FlashInfer, SGLang's kernels), CUDA 13,
+// x86-64 only. A node of older cards is not one it is meant for, and is not
+// reported as a fault; one of new cards without the toolkit is.
+func freetokenCandidates(version, variant string, h discovery.Host) ([]Build, error) {
+	if _, err := gpuArch("FreeToken", h); err != nil {
+		return nil, err
+	}
+	if h.Arch != "amd64" {
+		return nil, NotHere{"FreeToken publishes wheels for x86-64 only; this node is " + h.Arch}
+	}
+	if cc := h.MinComputeCapability; cc > 0 && cc < freetokenMinCompute {
+		return nil, NotHere{fmt.Sprintf("FreeToken runs on GPUs of compute capability %.1f and newer "+
+			"(RTX 30, A100 and later); this node has %.1f", freetokenMinCompute, cc)}
+	}
+	// It compiles kernels the first time a model loads, with nvcc, as its
+	// own install guide says; without a toolkit the installation looks
+	// finished and the first replica fails.
+	if freetokenToolkit() == "" {
+		return nil, fmt.Errorf("FreeToken compiles its kernels where it runs and needs the CUDA 13 toolkit's nvcc " +
+			"(in CUDA_HOME, /usr/local/cuda or /usr/local/cuda-13.*); install cuda-toolkit-13 from NVIDIA's repository")
+	}
+	b := Build{
+		Variant:      "cu13",
+		Python:       pythonForEngines,
+		Requirements: []string{"freetoken[accel]==" + version},
+		Binary:       "ft",
+		Packages:     compilerPackages,
+		FreeGB:       20,
+		GPUCheck:     true,
+	}
+	return pick("FreeToken "+version, variant, []candidate{
+		{b, gpuNeeds{minDriverForCUDA13, freetokenMinCompute}, "FreeToken " + version},
+	}, h)
+}

internal/agent/provision/strata_test.go

@@ -0,0 +1,118 @@
+package provision
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/agent/discovery"
+)
+
+// strataRelease is Strata's source archive as GitHub makes one, and its
+// release's engine for CUDA 13, with the recipe pointed at the archive.
+func strataRelease(t *testing.T, rel *release) {
+	t.Helper()
+	archive := tarGz(t, map[string]string{
+		"Strata-0.2.7/setup.py":         "print('setup')\n",
+		"Strata-0.2.7/serve/server.py":  "print('server')\n",
+		"Strata-0.2.7/requirements.txt": "numpy==2.5.3; python_version >= \"3.12\"\njinja2==3.1.6\n",
+	}, nil)
+	rel.assets["v0.2.7.tar.gz"] = archive
+	sum := sha256.Sum256(archive)
+	engine := tarGz(t, map[string]string{"strata": "\x7fELF", "libcudart.so.13": "lib"}, nil)
+	name := "strata-engine-0.1.52-linux-x86_64-cuda-cuda13.tar.gz"
+	rel.assets[name] = engine
+	esum := sha256.Sum256(engine)
+	rel.digests[name] = "sha256:" + hex.EncodeToString(esum[:])
+
+	old := strataReleases
+	strataReleases = map[string]struct {
+		sha256    string
+		published string
+		engine    string
+	}{"0.2.7": {hex.EncodeToString(sum[:]), "2026-10-27T14:54:42Z", "0.1.52"}}
+	t.Cleanup(func() { strataReleases = old })
+}
+
+func TestStrataIsInstalledWithItsEngineBesideItsCode(t *testing.T) {
+	sys := &fakePython{fakeSystem: fakeSystem{installed: map[string]bool{}}}
+	in, rel := pythonInstaller(t, turingNode, sys)
+	strataRelease(t, rel)
+
+	got := in.Converge(t.Context(), []Want{{Engine: "strata", Version: "0.2.7"}})
+	if got[0].State != Ready {
+		t.Fatalf("not ready: %+v", got[0])
+	}
+	dir := filepath.Join(in.Root, "strata", "0.2.7-cuda13")
+	if got[0].Binary != filepath.Join(dir, "bin", "python") {
+		t.Errorf("binary %s, want the environment's interpreter", got[0].Binary)
+	}
+	for _, f := range []string{"strata/setup.py", "strata/serve/server.py", "strata-engine/strata", "strata-engine/libcudart.so.13"} {
+		if _, err := os.Stat(filepath.Join(dir, f)); err != nil {
+			t.Errorf("%s is not in the environment: %v", f, err)
+		}
+	}
+	all := strings.Join(sys.commands, "\n")
+	for _, want := range []string{"--exclude-newer 2026-10-28T14:54:42Z", "jinja2==3.1.6"} {
+		if !strings.Contains(all, want) {
+			t.Errorf("the installation never used %q:\n%s", want, all)
+		}
+	}
+}
+
+func TestStrataRefusesAnEngineWhoseDigestDoesNotMatch(t *testing.T) {
+	sys := &fakePython{fakeSystem: fakeSystem{installed: map[string]bool{}}}
+	in, rel := pythonInstaller(t, turingNode, sys)
+	strataRelease(t, rel)
+	rel.digests["strata-engine-0.1.52-linux-x86_64-cuda-cuda13.tar.gz"] = "sha256:" + strings.Repeat("0", 64)
+
+	got := in.Converge(t.Context(), []Want{{Engine: "strata", Version: "0.2.7", Variant: "cuda13"}})
+	if got[0].State != Failed || !strings.Contains(got[0].Detail, "does not match the checksum") {
+		t.Fatalf("an engine that does not match its digest was not refused: %+v", got[0])
+	}
+}
+
+func TestStrataPicksTheEngineForTheCardsAndDriver(t *testing.T) {
+	for _, tc := range []struct {
+		name string
+		host discovery.Host
+		want []string
+		err  string
+	}{
+		{"turing on 580", turingNode, []string{"cuda13", "cuda12"}, ""},
+		{"volta", discovery.Host{Arch: "amd64", NVIDIADriver: "580.82.07", MinComputeCapability: 7.0}, []string{"cuda12"}, ""},
+		{"driver 575", discovery.Host{Arch: "amd64", NVIDIADriver: "575.51.03", MinComputeCapability: 8.6}, []string{"cuda12"}, ""},
+		{"driver 570", discovery.Host{Arch: "amd64", NVIDIADriver: "570.86.10", MinComputeCapability: 8.6}, nil, "575"},
+		{"no gpu", discovery.Host{Arch: "amd64"}, nil, "no"},
+		{"arm", discovery.Host{Arch: "arm64", NVIDIADriver: "580.82.07", MinComputeCapability: 9.0}, nil, "x86-64"},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			bs, err := strataCandidates("0.2.7", "", tc.host)
+			if tc.err != "" {
+				if err == nil || !strings.Contains(err.Error(), tc.err) {
+					t.Fatalf("got %v, %v; want an error saying %q", bs, err, tc.err)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatal(err)
+			}
+			var got []string
+			for _, b := range bs {
+				got = append(got, b.Variant)
+			}
+			if strings.Join(got, ",") != strings.Join(tc.want, ",") {
+				t.Errorf("builds %v, want %v", got, tc.want)
+			}
+		})
+	}
+}
+
+func TestStrataIsNotInAnOfflineBundle(t *testing.T) {
+	if _, err := Builds("strata", "0.2.7", "amd64"); err == nil {
+		t.Fatal("an offline bundle would carry Strata, which needs PyPI")
+	}
+}

internal/agent/removeweights_test.go

@@ -94,3 +94,35 @@ func TestWithNoCacheConfiguredNothingIsDeleted(t *testing.T) {
 		t.Errorf("an agent with no cache directory deleted something: %v", err)
 	}
 }
+
+// A model of several files is fetched into a directory of its own, and an
+// unused one goes whole: it was refused as not empty, and nothing was freed.
+func TestAnUnusedSetIsRemovedWhole(t *testing.T) {
+	dir := t.TempDir()
+	a := agentWithCache(t, dir)
+	set := weights.CachePathFor(dir, "https://mirror/coder/v1")
+	if err := os.MkdirAll(set, 0o755); err != nil {
+		t.Fatal(err)
+	}
+	write(t, filepath.Join(set, "model-00001-of-00002.gguf"))
+	write(t, filepath.Join(set, ".complete"))
+	// A link in the cache is taken away, not what it points at.
+	target := t.TempDir()
+	write(t, filepath.Join(target, "keep.gguf"))
+	link := weights.CachePathFor(dir, "https://mirror/linked/v1")
+	if err := os.Symlink(target, link); err != nil {
+		t.Fatal(err)
+	}
+
+	a.removeWeights([]string{set, link})
+
+	if _, err := os.Stat(set); !os.IsNotExist(err) {
+		t.Errorf("the set is still there: %v", err)
+	}
+	if _, err := os.Lstat(link); !os.IsNotExist(err) {
+		t.Errorf("the link is still there: %v", err)
+	}
+	if _, err := os.Stat(filepath.Join(target, "keep.gguf")); err != nil {
+		t.Errorf("what the link pointed at was deleted: %v", err)
+	}
+}

internal/agent/slurm.go

@@ -0,0 +1,38 @@
+package agent
+
+import (
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+)
+
+// FeatureSlurm is the agent keeping its node's part in Slurm (BACKLOG P).
+const FeatureSlurm = "slurm"
+
+// slurmMarker says llmgw started Slurm's daemons on this machine.
+var slurmMarker = "/var/lib/llmgw-agent/slurm"
+
+// setSlurmOut is where Slurm's status goes now: this session's queue, or
+// nowhere between sessions.
+func (a *Agent) setSlurmOut(out chan<- *llmgwv1.AgentMessage) {
+	a.mu.Lock()
+	a.slurmOut = out
+	a.mu.Unlock()
+	if out != nil {
+		a.slurm.Resend()
+	}
+}
+
+// sendSlurm queues a status without waiting: a full queue or no session
+// drops it, and the next session is sent it again.
+func (a *Agent) sendSlurm(st *llmgwv1.SlurmStatus) {
+	a.mu.Lock()
+	out := a.slurmOut
+	a.mu.Unlock()
+	if out == nil {
+		return
+	}
+	select {
+	case out <- &llmgwv1.AgentMessage{Msg: &llmgwv1.AgentMessage_SlurmStatus{SlurmStatus: st}}:
+	default:
+		a.log.Warn("Slurm's status not sent: the queue to the control plane is full")
+	}
+}

internal/agent/slurmnode/containers.go

@@ -0,0 +1,198 @@
+package slurmnode
+
+import (
+	"context"
+	"fmt"
+	"path/filepath"
+	"runtime"
+	"strings"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+// cacheDir holds what is fetched for Slurm: enroot's packages, pyxis's
+// source and its build.
+const cacheDir = "/var/cache/llmgw-agent/slurm"
+
+// pyxisMarker is the pyxis version built here, against this Slurm.
+func (k *Keeper) pyxisMarker() string { return k.marker + "-pyxis" }
+
+// containers has enroot and pyxis here, set up, when want says so; what
+// is there is left when it no longer does (Slurm stops loading pyxis once
+// the controller's plugstack.conf goes). Answers the status line, empty
+// while either is missing.
+func (k *Keeper) containers(ctx context.Context, want *llmgwv1.Slurm) (string, error) {
+	if !want.GetContainers() {
+		return "", nil
+	}
+	family := k.family()
+	if family == "" {
+		return "", fmt.Errorf("neither dpkg nor rpm here: enroot and pyxis come as deb or rpm")
+	}
+	if err := k.enroot(ctx, family); err != nil {
+		return "", fmt.Errorf("enroot: %w", err)
+	}
+	if err := k.enrootConf(ctx, want.GetShared().GetPath()); err != nil {
+		return "", fmt.Errorf("enroot's settings: %w", err)
+	}
+	if _, err := k.host.LookPath("nvidia-smi"); err == nil {
+		if err := k.nvidiaContainer(ctx, family); err != nil {
+			return "", fmt.Errorf("the GPUs in containers (nvidia-container-cli): %w", err)
+		}
+	}
+	if err := k.pyxis(ctx, family); err != nil {
+		return "", fmt.Errorf("pyxis: %w", err)
+	}
+	return slurm.ContainersStatus(), nil
+}
+
+func (k *Keeper) family() string {
+	if _, err := k.host.LookPath("dpkg"); err == nil {
+		return "deb"
+	}
+	if _, err := k.host.LookPath("rpm"); err == nil {
+		return "rpm"
+	}
+	return ""
+}
+
+// enroot installs the pinned release when another, or none, is here.
+func (k *Keeper) enroot(ctx context.Context, family string) error {
+	if k.installed["enroot"] {
+		return nil
+	}
+	if out, err := k.host.Run(ctx, nil, "enroot", "version"); err == nil && strings.TrimSpace(string(out)) == slurm.EnrootVersion {
+		k.installed["enroot"] = true
+		return nil
+	}
+	pkgs, err := slurm.EnrootPackages(family, runtime.GOARCH)
+	if err != nil {
+		return err
+	}
+	if k.inst == nil {
+		return fmt.Errorf("enroot missing, and this agent installs nothing")
+	}
+	var files []string
+	for _, p := range pkgs {
+		f := filepath.Join(cacheDir, p.Name)
+		if err := k.host.Download(ctx, p.URL(), f, p.SHA256); err != nil {
+			return err
+		}
+		files = append(files, f)
+	}
+	k.log.Info("installing enroot, for Slurm's jobs in containers", "version", slurm.EnrootVersion)
+	if err := k.inst.InstallPackages(ctx, map[string][]string{family: files}); err != nil {
+		return err
+	}
+	out, err := k.host.Run(ctx, nil, "enroot", "version")
+	if v := strings.TrimSpace(string(out)); err != nil || v != slurm.EnrootVersion {
+		return fmt.Errorf("enroot says %q after installing %s: %v", v, slurm.EnrootVersion, err)
+	}
+	k.installed["enroot"] = true
+	return nil
+}
+
+// enrootConf writes enroot's settings and makes its directories.
+func (k *Keeper) enrootConf(ctx context.Context, shared string) error {
+	if _, err := k.write(slurm.EnrootConfPath, []byte(slurm.EnrootConf), 0o644, ""); err != nil {
+		return err
+	}
+	changed, err := k.write(slurm.EnrootTmpfiles, []byte(slurm.EnrootTmpfilesConf), 0o644, "")
+	if err != nil {
+		return err
+	}
+	if changed {
+		if out, err := k.host.Run(ctx, nil, "systemd-tmpfiles", "--create", slurm.EnrootTmpfiles); err != nil {
+			return fmt.Errorf("systemd-tmpfiles: %v: %s", err, lastLine(out))
+		}
+	}
+	for _, d := range []string{slurm.EnrootLib + "/cache", slurm.EnrootLib + "/data"} {
+		if _, err := k.host.Run(ctx, nil, "test", "-d", d); err == nil {
+			continue
+		}
+		// Each user makes their own beneath, as in /tmp.
+		if out, err := k.host.Run(ctx, nil, "install", "-d", "-m", "1777", d); err != nil {
+			return fmt.Errorf("making %s: %v: %s", d, err, lastLine(out))
+		}
+	}
+	if shared == "" {
+		return k.host.Remove(slurm.EnrootMounts)
+	}
+	_, err = k.write(slurm.EnrootMounts, []byte(slurm.EnrootSharedMount(shared)), 0o644, "")
+	return err
+}
+
+// nvidiaContainer installs nvidia-container-cli, which enroot's hook uses
+// to give a container the job's GPUs, from NVIDIA's repository.
+func (k *Keeper) nvidiaContainer(ctx context.Context, family string) error {
+	if k.installed["nvidia-container-cli"] {
+		return nil
+	}
+	if _, err := k.host.LookPath("nvidia-container-cli"); err == nil {
+		k.installed["nvidia-container-cli"] = true
+		return nil
+	}
+	if k.inst == nil {
+		return fmt.Errorf("nvidia-container-cli missing, and this agent installs nothing")
+	}
+	switch family {
+	case "deb":
+		if err := k.host.Download(ctx, slurm.NvidiaContainerKeyURL, slurm.NvidiaContainerKeyPath,
+			slurm.NvidiaContainerKeySHA256); err != nil {
+			return err
+		}
+		if _, err := k.write(slurm.NvidiaContainerAptList, []byte(slurm.NvidiaContainerApt), 0o644, ""); err != nil {
+			return err
+		}
+	case "rpm":
+		if _, err := k.write(slurm.NvidiaContainerYumRepo, []byte(slurm.NvidiaContainerYum), 0o644, ""); err != nil {
+			return err
+		}
+	}
+	k.log.Info("installing nvidia-container-cli, so containers get the job's GPUs")
+	if err := k.inst.InstallPackages(ctx, slurm.NvidiaContainerPackages); err != nil {
+		return err
+	}
+	if _, err := k.host.LookPath("nvidia-container-cli"); err != nil {
+		return fmt.Errorf("still missing after installing libnvidia-container-tools")
+	}
+	k.installed["nvidia-container-cli"] = true
+	return nil
+}
+
+// pyxis builds the pinned pyxis against this machine's Slurm and installs
+// it, when it has not been, or was for another version.
+func (k *Keeper) pyxis(ctx context.Context, family string) error {
+	if b, err := k.host.ReadFile(k.pyxisMarker()); err == nil && strings.TrimSpace(string(b)) == slurm.PyxisVersion {
+		if _, err := k.host.Run(ctx, nil, "test", "-f", slurm.PyxisPlugin); err == nil {
+			return nil
+		}
+	}
+	if k.inst == nil {
+		return fmt.Errorf("pyxis missing, and this agent installs nothing")
+	}
+	if err := k.inst.InstallPackages(ctx, map[string][]string{family: slurm.PyxisBuildPackages[family]}); err != nil {
+		return fmt.Errorf("installing what builds it: %w", err)
+	}
+	tarball := filepath.Join(cacheDir, "pyxis-"+slurm.PyxisVersion+".tar.gz")
+	if err := k.host.Download(ctx, slurm.PyxisSource, tarball, slurm.PyxisSHA256); err != nil {
+		return err
+	}
+	src := filepath.Join(cacheDir, "pyxis-"+slurm.PyxisVersion)
+	k.log.Info("building pyxis against this machine's Slurm", "version", slurm.PyxisVersion)
+	for _, c := range [][]string{
+		{"rm", "-rf", src},
+		{"tar", "-xzf", tarball, "-C", cacheDir},
+		{"make", "-C", src, "install"},
+	} {
+		if out, err := k.host.Run(ctx, nil, c[0], c[1:]...); err != nil {
+			return fmt.Errorf("%s: %v: %s", strings.Join(c, " "), err, lastLine(out))
+		}
+	}
+	if _, err := k.host.Run(ctx, nil, "test", "-f", slurm.PyxisPlugin); err != nil {
+		return fmt.Errorf("built, yet %s is not there", slurm.PyxisPlugin)
+	}
+	_, err := k.write(k.pyxisMarker(), []byte(slurm.PyxisVersion+"\n"), 0o644, "")
+	return err
+}

internal/agent/slurmnode/containers_test.go

@@ -0,0 +1,92 @@
+package slurmnode
+
+import (
+	"context"
+	"runtime"
+	"slices"
+	"strings"
+	"testing"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+func TestContainersInstalledFromPinnedReleases(t *testing.T) {
+	if _, err := slurm.EnrootPackages("deb", runtime.GOARCH); err != nil {
+		t.Skip(err)
+	}
+	h := newFake()
+	h.bins["dpkg"], h.bins["nvidia-smi"] = true, true
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/var/lib/llmgw-agent/slurm", nil)
+	// enroot is not here until installed, and then says its version.
+	h.fail["enroot version"] = true
+	h.fail["test -d "+slurm.EnrootLib+"/cache"] = true
+	h.fail["test -f "+slurm.PyxisPlugin] = true
+	m := nodeMsg(true)
+	m.Containers = true
+	m.Shared = &llmgwv1.SlurmShared{Path: "/shared", Nfs: "10.0.0.11:/shared"}
+
+	st := k.Reconcile(context.Background(), m)
+	if st.GetContainers() != "" || !strings.Contains(st.GetError(), "containers: enroot") {
+		t.Fatalf("ready with enroot saying nothing: %v", st)
+	}
+	pkgs, _ := slurm.EnrootPackages("deb", runtime.GOARCH)
+	for _, p := range pkgs {
+		if !slices.Contains(h.ran, "download "+p.URL()+" "+cacheDir+"/"+p.Name+" "+p.SHA256) {
+			t.Errorf("%s not fetched with its digest:\n%s", p.Name, strings.Join(h.ran, "\n"))
+		}
+	}
+	if len(inst.got["deb"]) != 2 || !strings.HasPrefix(inst.got["deb"][0], cacheDir) {
+		t.Errorf("installed %v", inst.got)
+	}
+
+	delete(h.fail, "enroot version")
+	h.outputs["enroot version"] = slurm.EnrootVersion + "\n"
+	h.bins["nvidia-container-cli"] = false
+	st = k.Reconcile(context.Background(), m)
+	if !strings.Contains(st.GetError(), "nvidia-container-cli") {
+		t.Errorf("a GPU machine without nvidia-container-cli is ready: %v", st)
+	}
+	if string(h.files[slurm.NvidiaContainerAptList]) != slurm.NvidiaContainerApt ||
+		!slices.Contains(h.ran, "download "+slurm.NvidiaContainerKeyURL+" "+slurm.NvidiaContainerKeyPath+" "+slurm.NvidiaContainerKeySHA256) {
+		t.Errorf("NVIDIA's repository not added with its key checked")
+	}
+
+	h.bins["nvidia-container-cli"] = true
+	st = k.Reconcile(context.Background(), m)
+	// Built, but the plugin is still not there.
+	if st.GetContainers() != "" || !strings.Contains(st.GetError(), "pyxis") {
+		t.Fatalf("status %v", st)
+	}
+	for _, want := range []string{"tar -xzf " + cacheDir + "/pyxis-" + slurm.PyxisVersion + ".tar.gz -C " + cacheDir,
+		"make -C " + cacheDir + "/pyxis-" + slurm.PyxisVersion + " install", "install -d -m 1777 " + slurm.EnrootLib + "/cache"} {
+		if !slices.Contains(h.ran, want) {
+			t.Errorf("did not run %q", want)
+		}
+	}
+	delete(h.fail, "test -f "+slurm.PyxisPlugin)
+	st = k.Reconcile(context.Background(), m)
+	if st.GetContainers() != slurm.ContainersStatus() || st.GetError() != "" {
+		t.Fatalf("status %v", st)
+	}
+	if string(h.files[slurm.EnrootConfPath]) != slurm.EnrootConf ||
+		!strings.Contains(string(h.files[slurm.EnrootMounts]), "/shared /shared none") {
+		t.Errorf("enroot's settings not written")
+	}
+
+	// Done: nothing fetched or built again.
+	h.ran = nil
+	k.Reconcile(context.Background(), m)
+	for _, c := range h.ran {
+		if strings.HasPrefix(c, "download") || strings.HasPrefix(c, "make") {
+			t.Errorf("ran %q again", c)
+		}
+	}
+
+	// Off: nothing said, nothing taken away.
+	m.Containers = false
+	if st := k.Reconcile(context.Background(), m); st.GetContainers() != "" {
+		t.Errorf("status %v", st)
+	}
+}

internal/agent/slurmnode/dbd.go

@@ -0,0 +1,170 @@
+package slurmnode
+
+import (
+	"bufio"
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"strings"
+	"time"
+
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+// mariaDBConfDirs are where each family's MariaDB reads its settings: Debian
+// and Ubuntu's, then Red Hat's.
+var mariaDBConfDirs = []string{"/etc/mysql/mariadb.conf.d", "/etc/my.cnf.d"}
+
+const mariaDBConfName = "90-llmgw-slurm.cnf"
+
+// acctSQLPath is where the database's statements are put for the client to
+// read, never on its command line: they hold the password.
+const acctSQLPath = "/run/llmgw-slurm-acct.sql"
+
+// accounting keeps slurmdbd on the controller: MariaDB on this machine
+// alone, the database and its user, slurmdbd on them, and the cluster in it.
+// Before slurmctld, which reports to it from its first start.
+func (k *Keeper) accounting(ctx context.Context, key []byte, files map[string]string) error {
+	password := slurm.DbdPassword(key)
+	changed, err := k.mariaDBSettings(ctx)
+	if err != nil {
+		return err
+	}
+	switch {
+	case changed:
+		k.log.Info("MariaDB, for Slurm's accounting, starts on this machine alone")
+		if err := k.systemctl(ctx, "enable", "--now", "mariadb", "&&restart"); err != nil {
+			return err
+		}
+	case !k.active(ctx, "mariadb"):
+		if err := k.systemctl(ctx, "enable", "--now", "mariadb"); err != nil {
+			return err
+		}
+	}
+	client, err := k.mariaDBClient()
+	if err != nil {
+		return err
+	}
+	if err := k.waitFor(ctx, "MariaDB", client, "-e", "SELECT 1"); err != nil {
+		return err
+	}
+	// The database and its user, once each time the agent starts or the
+	// password is new: cheap, and a database restored from elsewhere gets
+	// its user back.
+	sum := sha256.Sum256([]byte(password))
+	if mark := hex.EncodeToString(sum[:]); k.acctDone != mark {
+		if err := k.host.WriteFile(acctSQLPath, []byte(slurm.AcctSQL(password)), 0o600, ""); err != nil {
+			return err
+		}
+		out, err := k.host.Run(ctx, nil, client, "-e", "source "+acctSQLPath)
+		_ = k.host.Remove(acctSQLPath)
+		if err != nil {
+			return fmt.Errorf("making Slurm's database: %v: %s", err, lastLine(out))
+		}
+		k.acctDone = mark
+	}
+
+	out, err := k.host.Run(ctx, nil, "hostname", "-s")
+	host := strings.TrimSpace(string(out))
+	if err != nil || host == "" {
+		return fmt.Errorf("this machine's short hostname: %v: %s", err, lastLine(out))
+	}
+	changed, err = k.write(slurm.DbdConfPath, []byte(slurm.DbdConf(host, password)), 0o600, "slurm")
+	if err != nil {
+		return err
+	}
+	switch {
+	case changed:
+		k.log.Info("slurmdbd starts again on new settings")
+		if err := k.systemctl(ctx, "enable", "--now", "slurmdbd", "&&restart"); err != nil {
+			return err
+		}
+	case !k.active(ctx, "slurmdbd"):
+		if err := k.systemctl(ctx, "enable", "--now", "slurmdbd"); err != nil {
+			return err
+		}
+	}
+	if err := k.waitFor(ctx, "slurmdbd", "sacctmgr", "-n", "-P", "show", "cluster", "format=cluster"); err != nil {
+		return err
+	}
+	name := clusterName(files["slurm.conf"])
+	if name == "" || k.clusterAdded == name {
+		return nil
+	}
+	out, err = k.host.Run(ctx, nil, "sacctmgr", "-n", "-P", "show", "cluster", "format=cluster")
+	if err != nil {
+		return fmt.Errorf("sacctmgr show cluster: %v: %s", err, lastLine(out))
+	}
+	for _, l := range strings.Split(string(out), "\n") {
+		if strings.TrimSpace(l) == name {
+			k.clusterAdded = name
+			return nil
+		}
+	}
+	k.log.Info("the cluster is added to Slurm's accounting", "cluster", name)
+	if out, err := k.host.Run(ctx, nil, "sacctmgr", "-i", "add", "cluster", name); err != nil {
+		return fmt.Errorf("sacctmgr add cluster %s: %v: %s", name, err, lastLine(out))
+	}
+	k.clusterAdded = name
+	return nil
+}
+
+// mariaDBSettings keeps MariaDB to this machine, when llmgw installed it;
+// one that was here before llmgw keeps its own settings, and only gets the
+// database and its user.
+func (k *Keeper) mariaDBSettings(ctx context.Context) (bool, error) {
+	if _, err := k.host.ReadFile(k.mariaDBMarker()); err != nil {
+		return false, nil
+	}
+	for _, d := range mariaDBConfDirs {
+		if _, err := k.host.Run(ctx, nil, "test", "-d", d); err == nil {
+			return k.write(d+"/"+mariaDBConfName, []byte(slurm.MariaDBConf), 0o644, "")
+		}
+	}
+	return false, fmt.Errorf("MariaDB has none of %s to put its settings in", strings.Join(mariaDBConfDirs, ", "))
+}
+
+// mariaDBMarker is there when llmgw installed MariaDB on this machine.
+func (k *Keeper) mariaDBMarker() string { return k.marker + "-mariadb" }
+
+// mariaDBClient is MariaDB's command-line client, under either of its
+// names.
+func (k *Keeper) mariaDBClient() (string, error) {
+	for _, c := range []string{"mariadb", "mysql"} {
+		if _, err := k.host.LookPath(c); err == nil {
+			return c, nil
+		}
+	}
+	return "", errors.New("MariaDB's client (mariadb or mysql) is not installed")
+}
+
+// waitFor runs a command until it succeeds, half a minute at most.
+func (k *Keeper) waitFor(ctx context.Context, what, name string, args ...string) error {
+	var last []byte
+	for range 30 {
+		out, err := k.host.Run(ctx, nil, name, args...)
+		if err == nil {
+			return nil
+		}
+		last = out
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-time.After(pingWait):
+		}
+	}
+	return fmt.Errorf("%s did not answer after starting: %s", what, lastLine(last))
+}
+
+// clusterName is ClusterName in slurm.conf.
+func clusterName(conf string) string {
+	sc := bufio.NewScanner(strings.NewReader(conf))
+	for sc.Scan() {
+		if v, ok := strings.CutPrefix(strings.TrimSpace(sc.Text()), "ClusterName="); ok {
+			return strings.TrimSpace(v)
+		}
+	}
+	return ""
+}

internal/agent/slurmnode/shared.go

@@ -0,0 +1,150 @@
+package slurmnode
+
+import (
+	"context"
+	"fmt"
+	"path/filepath"
+	"strings"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+// What a machine says of the shared path.
+const (
+	SharedMounted  = "mounted"
+	SharedExported = "exported"
+	SharedMissing  = "missing"
+)
+
+// shared keeps the shared path as want says: exported from here, mounted
+// from an export, or looked for where the site mounts it. Answers how it
+// is.
+func (k *Keeper) shared(ctx context.Context, want *llmgwv1.Slurm) (string, error) {
+	sh := want.GetShared()
+	path := sh.GetPath()
+	k.unshare(ctx, path, sh.GetServe(), sh.GetNfs())
+	if path == "" {
+		return "", nil
+	}
+	if !strings.HasPrefix(path, "/") || strings.Contains(path, "..") || strings.ContainsAny(path, " \n\t\"'\\$`") {
+		return SharedMissing, fmt.Errorf("the leader sent a path %q", path)
+	}
+	switch {
+	case sh.GetServe():
+		if err := k.host.MkdirAll(path, 0o755, ""); err != nil {
+			return SharedMissing, err
+		}
+		changed, err := k.write(slurm.ExportsPath, []byte(slurm.Exports(path, sh.GetClients())), 0o644, "")
+		if err != nil {
+			return SharedMissing, err
+		}
+		if !k.active(ctx, "nfs-server") {
+			k.log.Info("NFS starts on this machine, exporting the Slurm cluster's shared files", "path", path)
+			if err := k.systemctl(ctx, "enable", "--now", "nfs-server"); err != nil {
+				return SharedMissing, err
+			}
+			changed = true
+		}
+		if changed {
+			if out, err := k.host.Run(ctx, nil, "exportfs", "-ra"); err != nil {
+				return SharedMissing, fmt.Errorf("exportfs -ra: %v: %s", err, lastLine(out))
+			}
+		}
+		return SharedExported, nil
+	case sh.GetNfs() != "":
+		nfs := sh.GetNfs()
+		if strings.ContainsAny(nfs, " \n\t\"'\\$`") || !strings.Contains(nfs, ":/") {
+			return SharedMissing, fmt.Errorf("the leader sent an export %q", nfs)
+		}
+		if err := k.host.MkdirAll(path, 0o755, ""); err != nil {
+			return SharedMissing, err
+		}
+		unit := slurm.MountUnitName(path)
+		changed, err := k.write(filepath.Join(slurm.MountUnitDir, unit), []byte(slurm.MountUnit(nfs, path)), 0o644, "")
+		if err != nil {
+			return SharedMissing, err
+		}
+		if changed {
+			k.log.Info("the Slurm cluster's shared files are mounted here", "from", nfs, "at", path)
+			if err := k.systemctl(ctx, "daemon-reload"); err != nil {
+				return SharedMissing, err
+			}
+			if err := k.systemctl(ctx, "enable", unit); err != nil {
+				return SharedMissing, err
+			}
+			// From another export: a restart mounts the new one, and fails
+			// while a job still has files open on the old.
+			if err := k.systemctl(ctx, "restart", unit); err != nil {
+				return SharedMissing, err
+			}
+		} else if !k.active(ctx, unit) {
+			if err := k.systemctl(ctx, "start", unit); err != nil {
+				return SharedMissing, err
+			}
+		}
+	}
+	if _, err := k.host.Run(ctx, nil, "mountpoint", "-q", path); err != nil {
+		if sh.GetNfs() == "" {
+			return SharedMissing, fmt.Errorf("%s is not mounted here: the site mounts it (slurm.shared names no export)", path)
+		}
+		return SharedMissing, fmt.Errorf("%s is not mounted from %s", path, sh.GetNfs())
+	}
+	return SharedMounted, nil
+}
+
+// unshare takes away the export or the mount llmgw kept here that want no
+// longer has: path is where the files are wanted now, serve and nfs how.
+func (k *Keeper) unshare(ctx context.Context, path string, serve bool, nfs string) {
+	if _, err := k.host.ReadFile(slurm.ExportsPath); err == nil && (!serve || path == "") {
+		k.log.Info("this machine no longer exports the Slurm cluster's shared files")
+		_ = k.host.Remove(slurm.ExportsPath)
+		_, _ = k.host.Run(ctx, nil, "exportfs", "-ra")
+	}
+	out, err := k.host.Run(ctx, nil, "systemctl", "list-units", "--all", "--plain", "--no-legend", "--type=mount")
+	if err != nil {
+		return
+	}
+	keep := ""
+	if path != "" && nfs != "" && !serve {
+		keep = slurm.MountUnitName(path)
+	}
+	for _, l := range strings.Split(string(out), "\n") {
+		f := strings.Fields(l)
+		if len(f) == 0 || f[0] == keep {
+			continue
+		}
+		file := filepath.Join(slurm.MountUnitDir, f[0])
+		b, err := k.host.ReadFile(file)
+		if err != nil || !strings.HasPrefix(string(b), "# Written by llmgw (BACKLOG P): the Slurm cluster's shared files.") {
+			continue
+		}
+		k.log.Info("the Slurm cluster's old shared mount is taken away", "unit", f[0])
+		if err := k.systemctl(ctx, "disable", "--now", f[0]); err != nil {
+			k.log.Warn("could not unmount the old shared files", "unit", f[0], "err", err)
+			continue
+		}
+		_ = k.host.Remove(file)
+		_ = k.systemctl(ctx, "daemon-reload")
+	}
+}
+
+// sharedPackages are what the shared path needs here, by the command that
+// shows each is installed.
+func sharedPackages(want *llmgwv1.Slurm, need map[string][2]string) {
+	switch sh := want.GetShared(); {
+	case sh.GetPath() == "":
+	case sh.GetServe():
+		need["exportfs"] = [2]string{"nfs-kernel-server", "nfs-utils"}
+	case sh.GetNfs() != "":
+		need["mount.nfs"] = [2]string{"nfs-common", "nfs-utils"}
+	}
+}
+
+// Serves is whether this machine exports the shared files, so NFS's port
+// is kept for the cluster's networks.
+func (k *Keeper) Serves() bool {
+	k.mu.Lock()
+	defer k.mu.Unlock()
+	return k.want.GetShared().GetServe() && k.want.GetShared().GetPath() != ""
+}

internal/agent/slurmnode/shared_test.go

@@ -0,0 +1,171 @@
+package slurmnode
+
+import (
+	"context"
+	"slices"
+	"strings"
+	"testing"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+func ayse() *llmgwv1.SlurmUser {
+	return &llmgwv1.SlurmUser{Login: "ayse", Uid: 200000, Name: "Ayşe Y.", Home: "/shared/home/ayse",
+		SshKeys: []string{"ssh-ed25519 AAAA ayse@laptop"}}
+}
+
+func TestTheServerExportsAndMakesUsersWithHomes(t *testing.T) {
+	h := newFake()
+	h.active["ssh"] = true
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/var/lib/llmgw-agent/slurm", nil)
+	m := nodeMsg(false)
+	m.Shared = &llmgwv1.SlurmShared{Path: "/shared", Serve: true, Clients: []string{"10.0.0.22", "10.0.0.21"}}
+	m.ManageUsers, m.MakeHomes, m.Users = true, true, []*llmgwv1.SlurmUser{ayse()}
+	h.fail["test -d /shared/home/ayse"] = true
+	h.files[slurm.PAMSSHD] = []byte(pamStack)
+	st := k.Reconcile(context.Background(), m)
+	if st.GetError() != "" {
+		t.Fatal(st.GetError())
+	}
+	if st.GetShared() != SharedExported || st.GetUsers() != 1 {
+		t.Fatalf("status %v", st)
+	}
+	if !slices.Contains(inst.got["deb"], "nfs-kernel-server") {
+		t.Errorf("packages %v", inst.got)
+	}
+	if e := string(h.files[slurm.ExportsPath]); !strings.Contains(e, `"/shared" 10.0.0.21(`) {
+		t.Errorf("exports:\n%s", e)
+	}
+	for _, want := range []string{"systemctl enable --now nfs-server", "exportfs -ra", "groupadd -g 200000 ayse",
+		"useradd -u 200000 -g 200000 -M -d /shared/home/ayse -s /bin/bash -c Ayşe Y. ayse",
+		"install -d -m 0700 -o 200000 -g 200000 /shared/home/ayse", "sshd -t", "systemctl reload ssh"} {
+		if !slices.Contains(h.ran, want) {
+			t.Errorf("did not run %q:\n%s", want, strings.Join(h.ran, "\n"))
+		}
+	}
+	if string(h.files[slurm.AuthorizedKeysDir+"/ayse"]) != "ssh-ed25519 AAAA ayse@laptop\n" ||
+		string(h.files[k.usersMarker()]) != "ayse\n" || h.files[slurm.SSHDropIn] == nil {
+		t.Errorf("files: keys %q, marker %q", h.files[slurm.AuthorizedKeysDir+"/ayse"], h.files[k.usersMarker()])
+	}
+
+	// Again, made: nothing is run again.
+	h.outputs["getent passwd ayse"] = "ayse:x:200000:200000:Ayşe Y.:/shared/home/ayse:/bin/bash\n"
+	delete(h.fail, "test -d /shared/home/ayse")
+	h.ran = nil
+	k.Reconcile(context.Background(), m)
+	for _, c := range h.ran {
+		if strings.HasPrefix(c, "useradd") || strings.HasPrefix(c, "usermod") || strings.HasPrefix(c, "exportfs") ||
+			strings.HasPrefix(c, "install") || c == "systemctl reload ssh" {
+			t.Errorf("ran %q again", c)
+		}
+	}
+
+	// Gone from the accounts: taken away, its home left.
+	m.Users = nil
+	k.Reconcile(context.Background(), m)
+	if !slices.Contains(h.ran, "userdel ayse") || h.files[k.usersMarker()] != nil || h.files[slurm.SSHDropIn] != nil ||
+		h.files[slurm.AuthorizedKeysDir+"/ayse"] != nil {
+		t.Errorf("not taken away:\n%s", strings.Join(h.ran, "\n"))
+	}
+	if slices.ContainsFunc(h.ran, func(c string) bool { return strings.HasPrefix(c, "rm") || strings.Contains(c, "userdel -r") }) {
+		t.Error("a home was removed")
+	}
+}
+
+func TestAClientMountsAndLeavesOthersUsersAlone(t *testing.T) {
+	h := newFake()
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/var/lib/llmgw-agent/slurm", nil)
+	m := nodeMsg(false)
+	m.Shared = &llmgwv1.SlurmShared{Path: "/shared", Nfs: "10.0.0.11:/shared"}
+	bob := &llmgwv1.SlurmUser{Login: "bob", Uid: 200001, Home: "/shared/home/bob"}
+	m.ManageUsers, m.Users = true, []*llmgwv1.SlurmUser{ayse(), bob}
+	// The site's own ayse, with its own UID.
+	h.outputs["getent passwd ayse"] = "ayse:x:1001:1001::/home/ayse:/bin/bash\n"
+	st := k.Reconcile(context.Background(), m)
+	if st.GetShared() != SharedMounted || st.GetUsers() != 1 {
+		t.Fatalf("status %v", st)
+	}
+	if !strings.Contains(st.GetError(), "ayse is already a user here with UID 1001") {
+		t.Errorf("error %q", st.GetError())
+	}
+	if u := string(h.files["/etc/systemd/system/shared.mount"]); !strings.Contains(u, "What=10.0.0.11:/shared") {
+		t.Fatalf("unit:\n%s", u)
+	}
+	for _, want := range []string{"systemctl enable shared.mount", "systemctl restart shared.mount", "mountpoint -q /shared",
+		"useradd -u 200001 -g 200001 -M -d /shared/home/bob -s /bin/bash -c  bob"} {
+		if !slices.Contains(h.ran, want) {
+			t.Errorf("did not run %q:\n%s", want, strings.Join(h.ran, "\n"))
+		}
+	}
+	if slices.ContainsFunc(h.ran, func(c string) bool { return strings.Contains(c, "ayse") && !strings.HasPrefix(c, "getent") }) ||
+		slices.ContainsFunc(h.ran, func(c string) bool { return strings.HasPrefix(c, "install -d") }) {
+		t.Errorf("touched the site's ayse, or made a home:\n%s", strings.Join(h.ran, "\n"))
+	}
+	if string(h.files[k.usersMarker()]) != "bob\n" {
+		t.Errorf("marker %q", h.files[k.usersMarker()])
+	}
+
+	// Not mounted: said.
+	h.fail["mountpoint -q /shared"] = true
+	if st := k.Reconcile(context.Background(), m); st.GetShared() != SharedMissing ||
+		!strings.Contains(st.GetError(), "not mounted from 10.0.0.11:/shared") {
+		t.Errorf("status %v", st)
+	}
+
+	// No shared path any more: the mount is taken away.
+	h.outputs["systemctl list-units --all --plain --no-legend --type=mount"] = "shared.mount loaded active mounted llmgw\n"
+	m.Shared = nil
+	k.Reconcile(context.Background(), m)
+	if !slices.Contains(h.ran, "systemctl disable --now shared.mount") || h.files["/etc/systemd/system/shared.mount"] != nil {
+		t.Errorf("mount kept:\n%s", strings.Join(h.ran, "\n"))
+	}
+}
+
+const pamStack = "@include common-auth\naccount required pam_nologin.so\n@include common-account\n@include common-session\n"
+
+func TestUsersReachANodeOnlyThroughTheirJobs(t *testing.T) {
+	h := newFake()
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/var/lib/llmgw-agent/slurm", nil)
+	h.files[slurm.PAMSSHD] = []byte(pamStack)
+	// The module is not there until its package is installed.
+	for _, d := range slurm.ModuleDirs {
+		h.fail["test -f "+d+"/"+slurm.AdoptModule] = true
+	}
+	m := nodeMsg(true)
+	m.ManageUsers, m.Users = true, []*llmgwv1.SlurmUser{ayse()}
+	k.Reconcile(context.Background(), m)
+	if !slices.Contains(inst.got["deb"], "libpam-slurm-adopt") {
+		t.Fatalf("not installed: %v", inst.got)
+	}
+	if got := string(h.files[slurm.PAMSSHD]); got != pamStack {
+		t.Fatalf("lines put in with the module still missing:\n%s", got)
+	}
+	delete(h.fail, "test -f "+slurm.ModuleDirs[0]+"/"+slurm.AdoptModule)
+	st := k.Reconcile(context.Background(), m)
+	got := string(h.files[slurm.PAMSSHD])
+	if !strings.Contains(got, "pam_listfile.so item=user sense=allow file="+k.usersMarker()) ||
+		!strings.Contains(got, "account required "+slurm.AdoptModule) {
+		t.Fatalf("not held to their jobs (%s):\n%s", st.GetError(), got)
+	}
+
+	// The controller is where they log in: not held there.
+	m.Controller = true
+	m.Files = map[string]string{"slurm.conf": "x"}
+	k.Reconcile(context.Background(), m)
+	if string(h.files[slurm.PAMSSHD]) != pamStack {
+		t.Errorf("held on the controller:\n%s", h.files[slurm.PAMSSHD])
+	}
+
+	// No users of llmgw's any more: the lines go.
+	m.Controller, m.Files = false, nil
+	k.Reconcile(context.Background(), m)
+	m.Users = nil
+	k.Reconcile(context.Background(), m)
+	if string(h.files[slurm.PAMSSHD]) != pamStack {
+		t.Errorf("lines left with no users:\n%s", h.files[slurm.PAMSSHD])
+	}
+}

internal/agent/slurmnode/slurmnode.go

@@ -0,0 +1,733 @@
+// Package slurmnode keeps this machine's part in the Slurm cluster run beside
+// llmgw (BACKLOG P), as the leader says: slurmctld on the controller, slurmd
+// on a node, munge on both, installed from the machine's own package sources.
+//
+// What the leader sends is wanted state, kept by Keeper's loop: sent again,
+// or checked again a minute later, it changes nothing that is already so.
+package slurmnode
+
+import (
+	"bytes"
+	"context"
+	"crypto/sha256"
+	"encoding/hex"
+	"errors"
+	"fmt"
+	"io"
+	"io/fs"
+	"log/slog"
+	"net/http"
+	"os"
+	"os/exec"
+	"os/user"
+	"path/filepath"
+	"slices"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"google.golang.org/protobuf/proto"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+// Host is what Keeper does to the machine; Local in the agent, a fake in
+// tests.
+type Host interface {
+	Run(ctx context.Context, env []string, name string, args ...string) ([]byte, error)
+	LookPath(name string) (string, error)
+	ReadFile(path string) ([]byte, error)
+	// WriteFile writes b at path with mode, owned by owner ("" for root),
+	// making its directory.
+	WriteFile(path string, b []byte, mode fs.FileMode, owner string) error
+	// MkdirAll makes path, owned by owner.
+	MkdirAll(path string, mode fs.FileMode, owner string) error
+	Remove(path string) error
+	// Download fetches url to path, kept only if its SHA-256 is sha; a file
+	// already there with it is not fetched again.
+	Download(ctx context.Context, url, path, sha string) error
+}
+
+// Installer installs packages by family, "deb" and "rpm".
+type Installer interface {
+	InstallPackages(ctx context.Context, byFamily map[string][]string) error
+}
+
+// Reasons llmgw drains a node with; one drained for another reason, by a
+// Slurm administrator, is never resumed by llmgw.
+const (
+	ReasonServing = "llmgw: serving models"
+	reasonPrefix  = "llmgw"
+)
+
+// DropIn is the unit drop-in that starts slurmd configless.
+const DropIn = "/etc/systemd/system/slurmd.service.d/llmgw.conf"
+
+// Keeper keeps the machine as the last Slurm message says.
+type Keeper struct {
+	host   Host
+	inst   Installer
+	log    *slog.Logger
+	marker string
+
+	mu   sync.Mutex
+	want *llmgwv1.Slurm
+	kick chan struct{}
+	// resend has the next status sent even if unchanged: a new stream's
+	// control plane may never have had it.
+	resend bool
+	// installed is true once the packages were found or put there, so a
+	// minute's check does not look again.
+	installed map[string]bool
+	files     string
+	// acctDone is the digest of the password the database's user was last
+	// given; clusterAdded the cluster last found in the accounting.
+	acctDone     string
+	clusterAdded string
+}
+
+// New is a Keeper; marker is a file it keeps while it runs Slurm here, so
+// an agent started again knows to stop what it started when told it has no
+// part any more.
+func New(host Host, inst Installer, marker string, log *slog.Logger) *Keeper {
+	if log == nil {
+		log = slog.Default()
+	}
+	return &Keeper{host: host, inst: inst, log: log, marker: marker,
+		kick: make(chan struct{}, 1), installed: map[string]bool{}}
+}
+
+// Apply is a new message from the leader: kept from now on.
+func (k *Keeper) Apply(m *llmgwv1.Slurm) {
+	k.mu.Lock()
+	k.want = proto.CloneOf(m)
+	k.mu.Unlock()
+	select {
+	case k.kick <- struct{}{}:
+	default:
+	}
+}
+
+// HasPart is whether the last message gives this machine a part in Slurm.
+func (k *Keeper) HasPart() bool {
+	k.mu.Lock()
+	defer k.mu.Unlock()
+	return k.want.GetController() || k.want.GetNode()
+}
+
+// Resend has the status sent again on the next pass, which it starts now.
+func (k *Keeper) Resend() {
+	k.mu.Lock()
+	k.resend = true
+	k.mu.Unlock()
+	select {
+	case k.kick <- struct{}{}:
+	default:
+	}
+}
+
+// Every is how often the machine is checked again without a new message.
+var Every = time.Minute
+
+// resendAfter is how often an unchanged status is sent anyway.
+const resendAfter = 5 * time.Minute
+
+// Run keeps the machine until ctx ends, sending its status when it changes
+// and every few minutes while it has a part.
+func (k *Keeper) Run(ctx context.Context, send func(*llmgwv1.SlurmStatus)) {
+	t := time.NewTicker(Every)
+	defer t.Stop()
+	var last *llmgwv1.SlurmStatus
+	var sent time.Time
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-k.kick:
+		case <-t.C:
+		}
+		k.mu.Lock()
+		want, again := k.want, k.resend
+		k.resend = false
+		k.mu.Unlock()
+		if want == nil {
+			continue
+		}
+		st := k.Reconcile(ctx, want)
+		if st == nil {
+			last = nil
+			continue
+		}
+		if again || !proto.Equal(st, last) || time.Since(sent) > resendAfter {
+			send(st)
+			last, sent = st, time.Now()
+		}
+	}
+}
+
+// Reconcile makes the machine what want says once, and answers its status;
+// nil when it has no part in Slurm.
+func (k *Keeper) Reconcile(ctx context.Context, want *llmgwv1.Slurm) *llmgwv1.SlurmStatus {
+	ctx, cancel := context.WithTimeout(ctx, 20*time.Minute)
+	defer cancel()
+	if !want.GetController() && !want.GetNode() {
+		k.stop(ctx)
+		return nil
+	}
+	st := &llmgwv1.SlurmStatus{}
+	fail := func(what string, err error) *llmgwv1.SlurmStatus {
+		st.Error = what + ": " + err.Error()
+		k.log.Warn("keeping Slurm on this node: "+what, "err", err)
+		return st
+	}
+	if err := k.ensurePackages(ctx, want); err != nil {
+		return fail("installing Slurm", err)
+	}
+	// From the daemon, not scontrol: a client command looks for its files
+	// even to say its version, and a configless node has none before slurmd
+	// has fetched them.
+	daemon := "slurmd"
+	if !want.GetNode() {
+		daemon = "slurmctld"
+	}
+	out, err := k.host.Run(ctx, nil, daemon, "-V")
+	if err != nil {
+		return fail("reading Slurm's version", fmt.Errorf("%v: %s", err, lastLine(out)))
+	}
+	v, err := slurm.ParseVersion(string(out))
+	if err != nil {
+		return fail("reading Slurm's version", err)
+	}
+	st.Version = strings.TrimSpace(string(out))
+	if i := strings.LastIndexByte(st.Version, ' '); i >= 0 {
+		st.Version = st.Version[i+1:]
+	}
+	if slurm.TooOld(v) {
+		return fail("Slurm", fmt.Errorf("%d.%02d is older than %d.%02d, the first to run jobs in cgroup v2; "+
+			"a newer release of the distribution, or Slurm's own packages, has one", v[0], v[1],
+			slurm.MinVersion[0], slurm.MinVersion[1]))
+	}
+	if err := k.host.WriteFile(k.marker, []byte("llmgw runs Slurm here (BACKLOG P)\n"), 0o644, ""); err != nil {
+		return fail("keeping its marker", err)
+	}
+	if err := k.munge(ctx, want.GetMungeKey()); err != nil {
+		return fail("munge", err)
+	}
+	// The shared files, then the users whose homes are on them. Neither
+	// keeps Slurm from running: what goes wrong is said, and Slurm kept.
+	var shErr, usErr error
+	st.Shared, shErr = k.shared(ctx, want)
+	if shErr != nil {
+		k.log.Warn("keeping the Slurm cluster's shared files on this node", "err", shErr)
+	}
+	if shErr == nil || !want.GetMakeHomes() || want.GetShared().GetPath() == "" {
+		st.Users, usErr = k.users(ctx, want)
+	} else {
+		// Homes made now would be made beside the shared files, not on them.
+		usErr = errors.New("waiting for the shared files, where the homes are")
+	}
+	if usErr != nil {
+		k.log.Warn("keeping the Slurm users on this node", "err", usErr)
+	}
+	// Jobs in containers; Slurm runs without them, and is kept anyway.
+	var ctErr error
+	if st.Containers, ctErr = k.containers(ctx, want); ctErr != nil {
+		k.log.Warn("setting up jobs in containers on this node", "err", ctErr)
+	}
+	defer func() {
+		switch {
+		case st.Error != "":
+		case shErr != nil:
+			st.Error = "shared files: " + shErr.Error()
+		case usErr != nil:
+			st.Error = "users: " + usErr.Error()
+		case ctErr != nil:
+			st.Error = "containers: " + ctErr.Error()
+		}
+	}()
+	if want.GetController() {
+		// Accounting first: slurmctld reports to it from its first start.
+		// Without it slurmctld still runs jobs, a pass of backfill late
+		// each; so a failure here is said, and slurmctld is kept anyway.
+		var acctErr error
+		if acctErr = k.accounting(ctx, want.GetMungeKey(), want.GetFiles()); acctErr != nil {
+			k.log.Warn("keeping Slurm's accounting on this node", "err", acctErr)
+		}
+		st.SlurmdbdRunning = k.active(ctx, "slurmdbd")
+		defer func() {
+			if acctErr != nil && st.Error == "" {
+				st.Error = "slurmdbd: " + acctErr.Error()
+			}
+		}()
+		sum, err := k.controller(ctx, want.GetFiles())
+		st.FilesSha256 = sum
+		if err != nil {
+			return fail("slurmctld", err)
+		}
+		st.SlurmctldRunning = k.active(ctx, "slurmctld")
+	}
+	if want.GetNode() {
+		if err := k.node(ctx, want, st); err != nil {
+			return fail("slurmd", err)
+		}
+	}
+	return st
+}
+
+func (k *Keeper) ensurePackages(ctx context.Context, want *llmgwv1.Slurm) error {
+	need := map[string][2]string{
+		"munged":   {"munge", "munge"},
+		"scontrol": {"slurm-client", "slurm"},
+	}
+	if want.GetController() {
+		need["slurmctld"] = [2]string{"slurmctld", "slurm-slurmctld"}
+		need["slurmdbd"] = [2]string{"slurmdbd", "slurm-slurmdbd"}
+		need["mariadbd"] = [2]string{"mariadb-server", "mariadb-server"}
+	}
+	if want.GetNode() {
+		need["slurmd"] = [2]string{"slurmd", "slurm-slurmd"}
+	}
+	sharedPackages(want, need)
+	pkgs := map[string][]string{}
+	for bin, p := range need {
+		if k.installed[bin] {
+			continue
+		}
+		if _, err := k.host.LookPath(bin); err == nil {
+			k.installed[bin] = true
+			continue
+		}
+		pkgs["deb"] = append(pkgs["deb"], p[0])
+		pkgs["rpm"] = append(pkgs["rpm"], p[1])
+	}
+	if len(pkgs) == 0 {
+		return nil
+	}
+	slices.Sort(pkgs["deb"])
+	slices.Sort(pkgs["rpm"])
+	if k.inst == nil {
+		return fmt.Errorf("%s missing, and this agent installs nothing", strings.Join(pkgs["deb"], ", "))
+	}
+	k.log.Info("installing Slurm, as the cluster's settings say", "packages", pkgs["deb"])
+	if err := k.inst.InstallPackages(ctx, pkgs); err != nil {
+		return err
+	}
+	if slices.Contains(pkgs["deb"], "mariadb-server") {
+		// Installed by llmgw: its settings are llmgw's to keep. One that was
+		// here before is another's, and is left as it is.
+		if err := k.host.WriteFile(k.mariaDBMarker(), []byte("llmgw installed MariaDB here, for Slurm's accounting\n"),
+			0o644, ""); err != nil {
+			return err
+		}
+	}
+	for bin := range need {
+		if _, err := k.host.LookPath(bin); err != nil {
+			return fmt.Errorf("%s is still missing after installing %s", bin, strings.Join(pkgs["deb"], ", "))
+		}
+		k.installed[bin] = true
+	}
+	return nil
+}
+
+// munge writes the cluster's key and keeps munged running on it.
+func (k *Keeper) munge(ctx context.Context, key []byte) error {
+	if len(key) < 32 {
+		return errors.New("the leader sent no key")
+	}
+	changed, err := k.write(slurm.MungeKeyPath, key, 0o400, "munge")
+	if err != nil {
+		return err
+	}
+	if changed {
+		k.log.Info("munge key written; munge starts again on it")
+		return k.systemctl(ctx, "enable", "--now", "munge", "&&restart")
+	}
+	if !k.active(ctx, "munge") {
+		return k.systemctl(ctx, "enable", "--now", "munge")
+	}
+	return nil
+}
+
+// controller writes the files and keeps slurmctld running on them,
+// answering their digest.
+func (k *Keeper) controller(ctx context.Context, files map[string]string) (string, error) {
+	if files["slurm.conf"] == "" {
+		return "", errors.New("the leader sent no slurm.conf")
+	}
+	for _, d := range []string{slurm.StateDir, slurm.LogDir} {
+		if err := k.host.MkdirAll(d, 0o755, "slurm"); err != nil {
+			return "", err
+		}
+	}
+	names := make([]string, 0, len(files))
+	for n := range files {
+		names = append(names, n)
+	}
+	slices.Sort(names)
+	h := sha256.New()
+	changed := false
+	for _, n := range names {
+		if strings.ContainsAny(n, "/\\") || n == "" || n == "." || n == ".." {
+			return "", fmt.Errorf("the leader sent a file named %q", n)
+		}
+		fmt.Fprintf(h, "%s\x00%s\x00", n, files[n])
+		c, err := k.write(filepath.Join(slurm.ConfDir, n), []byte(files[n]), 0o644, "")
+		if err != nil {
+			return "", err
+		}
+		changed = changed || c
+	}
+	sum := hex.EncodeToString(h.Sum(nil))
+	if changed || sum != k.files {
+		if err := k.systemctl(ctx, "enable", "slurmctld"); err != nil {
+			return sum, err
+		}
+		// Nodes added or taken away need a restart before 23.11 and a
+		// reconfigure after it; a restart does both, and running jobs live
+		// through it.
+		k.log.Info("slurmctld starts again on new files")
+		if err := k.systemctl(ctx, "restart", "slurmctld"); err != nil {
+			return sum, err
+		}
+		if err := k.waitPing(ctx); err != nil {
+			return sum, err
+		}
+		// Hands the configless slurmd the new files.
+		if out, err := k.host.Run(ctx, nil, "scontrol", "reconfigure"); err != nil {
+			return sum, fmt.Errorf("scontrol reconfigure: %v: %s", err, lastLine(out))
+		}
+		k.files = sum
+		return sum, nil
+	}
+	if !k.active(ctx, "slurmctld") {
+		return sum, k.systemctl(ctx, "enable", "--now", "slurmctld")
+	}
+	return sum, nil
+}
+
+func (k *Keeper) waitPing(ctx context.Context) error {
+	var last []byte
+	for range 30 {
+		out, err := k.host.Run(ctx, nil, "scontrol", "ping")
+		if err == nil && bytes.Contains(out, []byte("UP")) {
+			return nil
+		}
+		last = out
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-time.After(pingWait):
+		}
+	}
+	return fmt.Errorf("slurmctld did not answer after starting: %s", lastLine(last))
+}
+
+var pingWait = time.Second
+
+// node keeps slurmd running configless and puts the node in or out of the
+// controller's hands as lent says.
+func (k *Keeper) node(ctx context.Context, want *llmgwv1.Slurm, st *llmgwv1.SlurmStatus) error {
+	name, server := want.GetName(), want.GetConfServer()
+	if name == "" || server == "" {
+		return errors.New("the leader sent no name or controller")
+	}
+	if strings.ContainsAny(name+server, " \n\t\"'\\$`") {
+		return fmt.Errorf("the leader sent a name or controller with spaces or quotes in it")
+	}
+	for _, d := range []string{slurm.SpoolDir, slurm.LogDir} {
+		if err := k.host.MkdirAll(d, 0o755, "slurm"); err != nil {
+			return err
+		}
+	}
+	bin, err := k.host.LookPath("slurmd")
+	if err != nil {
+		return err
+	}
+	unit := "# Written by llmgw (BACKLOG P): slurmd fetches its files from slurmctld.\n" +
+		"[Unit]\nConditionPathExists=\nStartLimitIntervalSec=0\n" +
+		// --systemd as the packaged unit has it (Type=notify, since 22.05):
+		// without it systemd waits for a readiness that never comes.
+		"[Service]\nExecStart=\nExecStart=" + bin + " --systemd --conf-server " + server + " -N " + name + "\n" +
+		// A node new to the controller is refused until the leader has
+		// written it into slurm.conf; tried again until it is.
+		"Restart=on-failure\nRestartSec=15\n"
+	changed, err := k.write(DropIn, []byte(unit), 0o644, "")
+	if err != nil {
+		return err
+	}
+	if out, err := k.host.Run(ctx, nil, "slurmd", "-C"); err == nil {
+		st.Hardware = strings.TrimSpace(string(out))
+	}
+	if changed {
+		if err := k.systemctl(ctx, "daemon-reload"); err != nil {
+			return err
+		}
+		if err := k.systemctl(ctx, "enable", "slurmd"); err != nil {
+			return err
+		}
+		if err := k.systemctl(ctx, "restart", "slurmd"); err != nil {
+			return err
+		}
+	} else if !k.active(ctx, "slurmd") {
+		if err := k.systemctl(ctx, "enable", "--now", "slurmd"); err != nil {
+			return err
+		}
+	}
+	st.SlurmdRunning = k.active(ctx, "slurmd")
+
+	st.Jobs = k.localJobs(ctx)
+	env := k.clientEnv()
+	out, err := k.host.Run(ctx, env, "scontrol", "show", "node", "-o", name)
+	if err != nil {
+		// Not in slurm.conf yet: the leader writes it once this node's
+		// hardware is in.
+		return nil
+	}
+	ns, err := slurm.ParseNodeState(string(out))
+	if err != nil {
+		return nil
+	}
+	ours := ns.Reason == "" || strings.HasPrefix(ns.Reason, reasonPrefix)
+	switch {
+	case want.GetLent() && ns.Drained() && ours:
+		k.log.Info("this node is lent to Slurm: resumed there")
+		if out, err := k.host.Run(ctx, env, "scontrol", "update", "NodeName="+name, "State=RESUME"); err != nil {
+			return fmt.Errorf("resuming the node: %v: %s", err, lastLine(out))
+		}
+	case !want.GetLent() && !strings.Contains(ns.State, "DRAIN"):
+		k.log.Info("this node is llmgw's: drained in Slurm, its jobs left to finish")
+		if out, err := k.host.Run(ctx, env, "scontrol", "update", "NodeName="+name, "State=DRAIN",
+			"Reason="+ReasonServing); err != nil {
+			return fmt.Errorf("draining the node: %v: %s", err, lastLine(out))
+		}
+	}
+	if out, err := k.host.Run(ctx, env, "scontrol", "show", "node", "-o", name); err == nil {
+		if ns2, err := slurm.ParseNodeState(string(out)); err == nil {
+			ns = ns2
+		}
+	}
+	st.State, st.Reason = ns.State, ns.Reason
+	return nil
+}
+
+// localJobs counts the job steps running here, from their slurmstepd, not
+// from the controller: a job outlives a controller that is down, and is
+// what decides whether llmgw may have the cards back. The one slurmstepd
+// that is no job's, "slurmstepd infinity" holding slurmd's cgroup v2 scope,
+// is not counted. -1 when it cannot tell.
+func (k *Keeper) localJobs(ctx context.Context) int32 {
+	// pgrep exits 1 with no output when there is none.
+	out, err := k.host.Run(ctx, nil, "pgrep", "-a", "-x", "slurmstepd")
+	if err != nil && len(bytes.TrimSpace(out)) > 0 {
+		return -1
+	}
+	n := int32(0)
+	for _, line := range strings.Split(strings.TrimSpace(string(out)), "\n") {
+		f := strings.Fields(line)
+		if len(f) < 2 || f[len(f)-1] == "infinity" {
+			continue
+		}
+		n++
+	}
+	return n
+}
+
+// clientEnv points scontrol at slurmd's copy of the files on a node that is
+// not the controller, which has none of its own.
+func (k *Keeper) clientEnv() []string {
+	if _, err := k.host.ReadFile(filepath.Join(slurm.ConfDir, "slurm.conf")); err == nil {
+		return nil
+	}
+	return []string{"SLURM_CONF=" + filepath.Join(slurm.SpoolDir, "conf-cache", "slurm.conf")}
+}
+
+// stop stops what llmgw started here, if it did.
+func (k *Keeper) stop(ctx context.Context) {
+	if _, err := k.host.ReadFile(k.marker); err != nil {
+		return
+	}
+	k.log.Info("this node has no part in Slurm any more: slurmd, slurmctld and slurmdbd stopped; the packages and MariaDB stay")
+	// MariaDB stays: it may have been the machine's before, and holds the
+	// jobs' records.
+	for _, u := range []string{"slurmd", "slurmctld", "slurmdbd"} {
+		_ = k.systemctl(ctx, "disable", "--now", u)
+	}
+	_ = k.host.Remove(DropIn)
+	_ = k.systemctl(ctx, "daemon-reload")
+	k.unshare(ctx, "", false, "")
+	if err := k.removeUsers(ctx, k.madeUsers(), nil); err != nil {
+		k.log.Warn("could not take away every Slurm user llmgw made", "err", err)
+	}
+	_ = k.host.Remove(k.marker)
+	k.files, k.acctDone, k.clusterAdded = "", "", ""
+}
+
+// write writes a file when it differs, saying whether it did.
+func (k *Keeper) write(path string, b []byte, mode fs.FileMode, owner string) (bool, error) {
+	if old, err := k.host.ReadFile(path); err == nil && bytes.Equal(old, b) {
+		return false, nil
+	}
+	return true, k.host.WriteFile(path, b, mode, owner)
+}
+
+// systemctl runs systemctl; a last argument "&&restart" restarts the unit
+// after.
+func (k *Keeper) systemctl(ctx context.Context, args ...string) error {
+	restart := false
+	if n := len(args); n > 0 && args[n-1] == "&&restart" {
+		args, restart = args[:n-1], true
+	}
+	if out, err := k.host.Run(ctx, nil, "systemctl", args...); err != nil {
+		return fmt.Errorf("systemctl %s: %v: %s", strings.Join(args, " "), err, lastLine(out))
+	}
+	if restart {
+		u := args[len(args)-1]
+		if out, err := k.host.Run(ctx, nil, "systemctl", "restart", u); err != nil {
+			return fmt.Errorf("systemctl restart %s: %v: %s", u, err, lastLine(out))
+		}
+	}
+	return nil
+}
+
+func (k *Keeper) active(ctx context.Context, unit string) bool {
+	out, err := k.host.Run(ctx, nil, "systemctl", "is-active", unit)
+	return err == nil && strings.TrimSpace(string(out)) == "active"
+}
+
+func lastLine(b []byte) string {
+	s := strings.TrimSpace(string(b))
+	if i := strings.LastIndexByte(s, '\n'); i >= 0 {
+		s = s[i+1:]
+	}
+	return s
+}
+
+// Local is the machine the agent runs on.
+type Local struct{}
+
+// Run runs a command, with env added to the agent's.
+func (Local) Run(ctx context.Context, env []string, name string, args ...string) ([]byte, error) {
+	cmd := exec.CommandContext(ctx, name, args...)
+	if len(env) > 0 {
+		cmd.Env = append(os.Environ(), env...)
+	}
+	return cmd.CombinedOutput()
+}
+
+// LookPath finds a command, in the sbin directories too: an agent started
+// by systemd has them, one started by hand may not.
+func (Local) LookPath(name string) (string, error) {
+	if p, err := exec.LookPath(name); err == nil {
+		return p, nil
+	}
+	// /usr/libexec: Red Hat's mariadbd.
+	for _, d := range []string{"/usr/sbin", "/sbin", "/usr/local/sbin", "/usr/libexec"} {
+		p := filepath.Join(d, name)
+		if fi, err := os.Stat(p); err == nil && !fi.IsDir() {
+			return p, nil
+		}
+	}
+	return "", fmt.Errorf("%s is not installed", name)
+}
+
+// ReadFile reads a file.
+func (Local) ReadFile(path string) ([]byte, error) { return os.ReadFile(path) } //nolint:gosec // our paths
+
+// WriteFile writes a file whole, through a temporary one beside it.
+func (Local) WriteFile(path string, b []byte, mode fs.FileMode, owner string) error {
+	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil { //nolint:gosec // system directories
+		return err
+	}
+	tmp := path + ".llmgw-new"
+	if err := os.WriteFile(tmp, b, mode); err != nil {
+		return err
+	}
+	if err := os.Chmod(tmp, mode); err != nil {
+		return err
+	}
+	if err := chown(tmp, owner); err != nil {
+		_ = os.Remove(tmp)
+		return err
+	}
+	return os.Rename(tmp, path)
+}
+
+// Download fetches url to path through a temporary file beside it, kept
+// only when its SHA-256 is sha.
+func (Local) Download(ctx context.Context, url, path, sha string) error {
+	if f, err := os.Open(path); err == nil { //nolint:gosec // our paths
+		h := sha256.New()
+		_, cerr := io.Copy(h, f)
+		_ = f.Close()
+		if cerr == nil && hex.EncodeToString(h.Sum(nil)) == sha {
+			return nil
+		}
+	}
+	if err := os.MkdirAll(filepath.Dir(path), 0o755); err != nil { //nolint:gosec // system directories
+		return err
+	}
+	ctx, cancel := context.WithTimeout(ctx, 10*time.Minute)
+	defer cancel()
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return err
+	}
+	resp, err := http.DefaultClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("fetching %s: %w", url, err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+	if resp.StatusCode != http.StatusOK {
+		return fmt.Errorf("fetching %s: %s", url, resp.Status)
+	}
+	tmp := path + ".llmgw-new"
+	f, err := os.OpenFile(tmp, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0o644) //nolint:gosec // our paths
+	if err != nil {
+		return err
+	}
+	h := sha256.New()
+	_, err = io.Copy(io.MultiWriter(f, h), io.LimitReader(resp.Body, 1<<30))
+	if cerr := f.Close(); err == nil {
+		err = cerr
+	}
+	if err != nil {
+		_ = os.Remove(tmp)
+		return fmt.Errorf("fetching %s: %w", url, err)
+	}
+	if got := hex.EncodeToString(h.Sum(nil)); got != sha {
+		_ = os.Remove(tmp)
+		return fmt.Errorf("%s has SHA-256 %s, not the %s llmgw was built with: not used", url, got, sha)
+	}
+	return os.Rename(tmp, path)
+}
+
+// MkdirAll makes a directory owned by owner.
+func (Local) MkdirAll(path string, mode fs.FileMode, owner string) error {
+	if err := os.MkdirAll(path, mode); err != nil {
+		return err
+	}
+	return chown(path, owner)
+}
+
+// Remove removes a file, and is content when there is none.
+func (Local) Remove(path string) error {
+	if err := os.Remove(path); err != nil && !errors.Is(err, fs.ErrNotExist) {
+		return err
+	}
+	return nil
+}
+
+func chown(path, owner string) error {
+	if owner == "" {
+		return nil
+	}
+	u, err := user.Lookup(owner)
+	if err != nil {
+		return fmt.Errorf("user %s: %w (installed with its package)", owner, err)
+	}
+	uid, _ := strconv.Atoi(u.Uid)
+	gid, _ := strconv.Atoi(u.Gid)
+	return os.Chown(path, uid, gid)
+}

internal/agent/slurmnode/slurmnode_test.go

@@ -0,0 +1,367 @@
+package slurmnode
+
+import (
+	"context"
+	"errors"
+	"io/fs"
+	"path/filepath"
+	"slices"
+	"strings"
+	"testing"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+type fakeHost struct {
+	files   map[string][]byte
+	owners  map[string]string
+	bins    map[string]bool
+	active  map[string]bool
+	ran     []string
+	outputs map[string]string
+	fail    map[string]bool
+	state   string
+	reason  string
+}
+
+func newFake() *fakeHost {
+	return &fakeHost{files: map[string][]byte{}, owners: map[string]string{}, bins: map[string]bool{},
+		active: map[string]bool{}, outputs: map[string]string{
+			"slurmd -V":     "slurm-wlm 23.11.4\n",
+			"slurmctld -V":  "slurm-wlm 23.11.4\n",
+			"scontrol ping": "Slurmctld(primary) at node06 is UP\n",
+			"slurmd -C":     "NodeName=cuda01 CPUs=16 RealMemory=64000\nUpTime=1-00:00:00\n",
+			"hostname -s":   "node06\n",
+		}, fail: map[string]bool{}, state: "IDLE+DRAIN", reason: slurm.ReasonNew}
+}
+
+func (f *fakeHost) Run(_ context.Context, env []string, name string, args ...string) ([]byte, error) {
+	c := strings.Join(append([]string{name}, args...), " ")
+	f.ran = append(f.ran, c)
+	if f.fail[c] {
+		return []byte("it failed"), errors.New("exit status 1")
+	}
+	switch {
+	case name == "systemctl" && len(args) >= 1:
+		u := args[len(args)-1]
+		switch args[0] {
+		case "is-active":
+			if f.active[u] {
+				return []byte("active\n"), nil
+			}
+			return []byte("inactive\n"), errors.New("exit status 3")
+		case "restart", "start":
+			f.active[u] = true
+		case "enable":
+			if slices.Contains(args, "--now") {
+				f.active[u] = true
+			}
+		case "disable":
+			f.active[u] = false
+		}
+		return []byte(f.outputs[c]), nil
+	case strings.HasPrefix(c, "scontrol show node"):
+		return []byte("NodeName=cuda01 Arch=x86_64 State=" + f.state + " Partitions=gpu Reason=" + f.reason + " [root@now]\n"), nil
+	case strings.HasPrefix(c, "scontrol update"):
+		if slices.Contains(args, "State=RESUME") {
+			f.state, f.reason = "IDLE", ""
+		} else {
+			f.state, f.reason = "IDLE+DRAIN", ReasonServing
+		}
+		return nil, nil
+	case strings.HasPrefix(c, "pgrep"):
+		out := "1800358 /usr/sbin/slurmstepd infinity\n" + f.outputs["jobs"]
+		return []byte(out), nil
+	}
+	return []byte(f.outputs[c]), nil
+}
+
+func (f *fakeHost) LookPath(name string) (string, error) {
+	if f.bins[name] {
+		return "/usr/sbin/" + name, nil
+	}
+	return "", errors.New("not found")
+}
+
+func (f *fakeHost) ReadFile(path string) ([]byte, error) {
+	b, ok := f.files[path]
+	if !ok {
+		return nil, fs.ErrNotExist
+	}
+	return b, nil
+}
+
+func (f *fakeHost) WriteFile(path string, b []byte, _ fs.FileMode, owner string) error {
+	f.files[path] = append([]byte(nil), b...)
+	f.owners[path] = owner
+	return nil
+}
+
+func (f *fakeHost) MkdirAll(string, fs.FileMode, string) error { return nil }
+
+func (f *fakeHost) Remove(path string) error { delete(f.files, path); return nil }
+
+func (f *fakeHost) Download(_ context.Context, url, path, sha string) error {
+	c := "download " + url + " " + path + " " + sha
+	f.ran = append(f.ran, c)
+	if f.fail[c] || f.fail["download "+url] {
+		return errors.New("fetch failed")
+	}
+	f.files[path] = []byte("fetched " + url)
+	return nil
+}
+
+func (f *fakeHost) count(prefix string) int {
+	n := 0
+	for _, c := range f.ran {
+		if strings.HasPrefix(c, prefix) {
+			n++
+		}
+	}
+	return n
+}
+
+type fakeInst struct {
+	host *fakeHost
+	got  map[string][]string
+}
+
+func (i *fakeInst) InstallPackages(_ context.Context, by map[string][]string) error {
+	i.got = by
+	for _, b := range []string{"munged", "scontrol", "slurmd", "slurmctld", "slurmdbd", "mariadbd", "mariadb", "exportfs", "mount.nfs"} {
+		i.host.bins[b] = true
+	}
+	return nil
+}
+
+var key = []byte(strings.Repeat("k", 1024))
+
+func nodeMsg(lent bool) *llmgwv1.Slurm {
+	return &llmgwv1.Slurm{Node: true, Name: "cuda01", ConfServer: "10.0.0.146:6817", MungeKey: key, Lent: lent}
+}
+
+func TestNodeInstallsAndStaysDrained(t *testing.T) {
+	h := newFake()
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/var/lib/llmgw-agent/slurm", nil)
+	st := k.Reconcile(context.Background(), nodeMsg(false))
+	if st.GetError() != "" {
+		t.Fatal(st.GetError())
+	}
+	if !slices.Equal(inst.got["deb"], []string{"munge", "slurm-client", "slurmd"}) ||
+		!slices.Equal(inst.got["rpm"], []string{"munge", "slurm", "slurm-slurmd"}) {
+		t.Fatalf("installed %v", inst.got)
+	}
+	if string(h.files[slurm.MungeKeyPath]) != string(key) || h.owners[slurm.MungeKeyPath] != "munge" {
+		t.Fatal("the munge key was not written for munge")
+	}
+	unit := string(h.files[DropIn])
+	if !strings.Contains(unit, "ExecStart=/usr/sbin/slurmd --systemd --conf-server 10.0.0.146:6817 -N cuda01\n") ||
+		!strings.Contains(unit, "ConditionPathExists=\n") {
+		t.Fatalf("drop-in:\n%s", unit)
+	}
+	if !st.GetSlurmdRunning() || st.GetVersion() != "23.11.4" || !strings.HasPrefix(st.GetHardware(), "NodeName=cuda01 CPUs=16") {
+		t.Fatalf("status %v", st)
+	}
+	// Not lent: left drained, with llmgw's first reason.
+	if h.count("scontrol update") != 0 || st.GetState() != "IDLE+DRAIN" {
+		t.Fatalf("a node not lent was changed: %v, %v", h.ran, st)
+	}
+	if _, ok := h.files["/var/lib/llmgw-agent/slurm"]; !ok {
+		t.Fatal("no marker")
+	}
+
+	// Again, nothing changed: no install, no restart.
+	h.ran = nil
+	inst.got = nil
+	k.Reconcile(context.Background(), nodeMsg(false))
+	if inst.got != nil || h.count("systemctl restart") != 0 || h.count("systemctl enable") != 0 {
+		t.Fatalf("a second pass did work: %v", h.ran)
+	}
+}
+
+func TestLendAndReclaim(t *testing.T) {
+	h := newFake()
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	st := k.Reconcile(context.Background(), nodeMsg(true))
+	if h.count("scontrol update NodeName=cuda01 State=RESUME") != 1 || st.GetState() != "IDLE" {
+		t.Fatalf("a lent node was not resumed: %v %v", h.ran, st)
+	}
+	h.outputs["jobs"] = "2001 slurmstepd: [101.batch]\n2002 slurmstepd: [102.0]\n"
+	st = k.Reconcile(context.Background(), nodeMsg(false))
+	if h.count("scontrol update NodeName=cuda01 State=DRAIN Reason="+ReasonServing) != 1 {
+		t.Fatalf("a node taken back was not drained: %v", h.ran)
+	}
+	if st.GetJobs() != 2 || !strings.Contains(st.GetState(), "DRAIN") {
+		t.Fatalf("status %v", st)
+	}
+	// Drained already: not drained again.
+	h.ran = nil
+	k.Reconcile(context.Background(), nodeMsg(false))
+	if h.count("scontrol update") != 0 {
+		t.Fatalf("drained twice: %v", h.ran)
+	}
+}
+
+func TestAdminsDrainIsLeft(t *testing.T) {
+	h := newFake()
+	h.reason = "bad DIMM, ticket 42"
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	k.Reconcile(context.Background(), nodeMsg(true))
+	if h.count("scontrol update") != 0 {
+		t.Fatalf("llmgw resumed a node a Slurm administrator drained: %v", h.ran)
+	}
+}
+
+func TestClientEnvOnANodeWithoutFiles(t *testing.T) {
+	h := newFake()
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	if env := k.clientEnv(); len(env) != 1 || !strings.HasSuffix(env[0], "conf-cache/slurm.conf") {
+		t.Fatalf("env %v", env)
+	}
+	h.files[filepath.Join(slurm.ConfDir, "slurm.conf")] = []byte("x")
+	if env := k.clientEnv(); env != nil {
+		t.Fatalf("env %v on the controller", env)
+	}
+}
+
+func TestController(t *testing.T) {
+	h := newFake()
+	inst := &fakeInst{host: h}
+	k := New(h, inst, "/m", nil)
+	files := map[string]string{"slurm.conf": "ClusterName=llmgw\n", "gres.conf": "#\n", "cgroup.conf": "#\n"}
+	m := &llmgwv1.Slurm{Controller: true, MungeKey: key, Files: files}
+	st := k.Reconcile(context.Background(), m)
+	if st.GetError() != "" {
+		t.Fatal(st.GetError())
+	}
+	if !slices.Contains(inst.got["deb"], "slurmctld") || slices.Contains(inst.got["deb"], "slurmd") {
+		t.Fatalf("installed %v", inst.got)
+	}
+	if string(h.files["/etc/slurm/slurm.conf"]) != "ClusterName=llmgw\n" {
+		t.Fatal("slurm.conf not written")
+	}
+	if h.count("systemctl restart slurmctld") != 1 || h.count("scontrol reconfigure") != 1 {
+		t.Fatalf("slurmctld not started on its files: %v", h.ran)
+	}
+	if !st.GetSlurmctldRunning() || len(st.GetFilesSha256()) != 64 {
+		t.Fatalf("status %v", st)
+	}
+	// Accounting: MariaDB on this machine alone, the database, slurmdbd on
+	// it, the cluster added, all before slurmctld.
+	if !slices.Contains(inst.got["deb"], "slurmdbd") || !slices.Contains(inst.got["deb"], "mariadb-server") {
+		t.Fatalf("installed %v", inst.got)
+	}
+	if !strings.Contains(string(h.files["/etc/mysql/mariadb.conf.d/90-llmgw-slurm.cnf"]), "bind-address=127.0.0.1") {
+		t.Fatal("MariaDB, installed by llmgw, not kept to this machine")
+	}
+	dbd := string(h.files[slurm.DbdConfPath])
+	if !strings.Contains(dbd, "DbdHost=node06\n") || !strings.Contains(dbd, "StoragePass="+slurm.DbdPassword(key)+"\n") ||
+		h.owners[slurm.DbdConfPath] != "slurm" {
+		t.Fatalf("slurmdbd.conf for %q:\n%s", h.owners[slurm.DbdConfPath], dbd)
+	}
+	if _, ok := h.files[acctSQLPath]; ok {
+		t.Fatal("the database's statements, password and all, were left on disk")
+	}
+	if strings.Contains(strings.Join(h.ran, "\n"), slurm.DbdPassword(key)) {
+		t.Fatal("the password was on a command line")
+	}
+	order := strings.Join(h.ran, "\n")
+	if !strings.Contains(order, "sacctmgr -i add cluster llmgw") ||
+		strings.Index(order, "slurmdbd") > strings.Index(order, "systemctl restart slurmctld") {
+		t.Fatalf("accounting not before slurmctld: %v", h.ran)
+	}
+	if !st.GetSlurmdbdRunning() {
+		t.Fatalf("status %v", st)
+	}
+	sum := st.GetFilesSha256()
+
+	h.ran = nil
+	k.Reconcile(context.Background(), m)
+	if h.count("systemctl restart") != 0 {
+		t.Fatalf("restarted on the same files: %v", h.ran)
+	}
+
+	files["gres.conf"] = "NodeName=cuda01 Name=gpu File=/dev/nvidia0\n"
+	h.ran = nil
+	st = k.Reconcile(context.Background(), m)
+	if h.count("systemctl restart slurmctld") != 1 || st.GetFilesSha256() == sum {
+		t.Fatalf("a changed file did not restart slurmctld: %v", h.ran)
+	}
+
+	if st := k.Reconcile(context.Background(), &llmgwv1.Slurm{Controller: true, MungeKey: key,
+		Files: map[string]string{"slurm.conf": "x", "../evil": "y"}}); !strings.Contains(st.GetError(), "named") {
+		t.Fatalf("a path in a file name was taken: %v", st)
+	}
+}
+
+func TestTooOld(t *testing.T) {
+	h := newFake()
+	h.outputs["slurmd -V"] = "slurm-wlm 21.08.5\n"
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	st := k.Reconcile(context.Background(), nodeMsg(false))
+	if !strings.Contains(st.GetError(), "older than 22.05") {
+		t.Fatalf("status %v", st)
+	}
+	if _, ok := h.files[slurm.MungeKeyPath]; ok {
+		t.Fatal("went on with a Slurm too old")
+	}
+}
+
+func TestStop(t *testing.T) {
+	h := newFake()
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	k.Reconcile(context.Background(), nodeMsg(false))
+	h.ran = nil
+	if st := k.Reconcile(context.Background(), &llmgwv1.Slurm{}); st != nil {
+		t.Fatalf("status %v for no part", st)
+	}
+	if !slices.Contains(h.ran, "systemctl disable --now slurmd") {
+		t.Fatalf("slurmd not stopped: %v", h.ran)
+	}
+	if _, ok := h.files[DropIn]; ok {
+		t.Fatal("drop-in left")
+	}
+	// Never started here: nothing touched.
+	h2 := newFake()
+	k2 := New(h2, &fakeInst{host: h2}, "/m", nil)
+	k2.Reconcile(context.Background(), &llmgwv1.Slurm{})
+	if len(h2.ran) != 0 {
+		t.Fatalf("a machine llmgw never ran Slurm on was touched: %v", h2.ran)
+	}
+}
+
+func TestBadNames(t *testing.T) {
+	h := newFake()
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	m := nodeMsg(false)
+	m.Name = "cuda01 -c /etc/evil"
+	if st := k.Reconcile(context.Background(), m); !strings.Contains(st.GetError(), "spaces") {
+		t.Fatalf("status %v", st)
+	}
+	if st := k.Reconcile(context.Background(), &llmgwv1.Slurm{Node: true, Name: "a", ConfServer: "b"}); !strings.Contains(st.GetError(), "no key") {
+		t.Fatalf("status %v", st)
+	}
+}
+
+func TestAMariaDBThatWasThereKeepsItsSettings(t *testing.T) {
+	h := newFake()
+	for _, b := range []string{"munged", "scontrol", "slurmctld", "slurmdbd", "mariadbd", "mariadb", "exportfs", "mount.nfs"} {
+		h.bins[b] = true
+	}
+	k := New(h, &fakeInst{host: h}, "/m", nil)
+	st := k.Reconcile(context.Background(), &llmgwv1.Slurm{Controller: true, MungeKey: key,
+		Files: map[string]string{"slurm.conf": "ClusterName=llmgw\n"}})
+	if st.GetError() != "" {
+		t.Fatal(st.GetError())
+	}
+	for p := range h.files {
+		if strings.HasSuffix(p, ".cnf") {
+			t.Fatalf("a MariaDB llmgw did not install had its settings changed: %s", p)
+		}
+	}
+	if h.count("mariadb -e source") != 1 || !st.GetSlurmdbdRunning() {
+		t.Fatalf("the database was not made in the MariaDB that was there: %v", h.ran)
+	}
+}

internal/agent/slurmnode/users.go

@@ -0,0 +1,319 @@
+package slurmnode
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"path/filepath"
+	"slices"
+	"strconv"
+	"strings"
+
+	llmgwv1 "gitea.mynodes.xyz/ilker/llmgw/api/gen/llmgw/v1"
+	"gitea.mynodes.xyz/ilker/llmgw/internal/slurm"
+)
+
+// usersMarker lists, a line each, the logins llmgw made here: the ones it
+// may change and take away. A user that was here before is never touched.
+func (k *Keeper) usersMarker() string { return k.marker + "-users" }
+
+// users keeps the Linux users want has, the same login and UID as on every
+// other machine, their keys where sshd reads them, and their homes when
+// this machine makes them; and takes away those llmgw made that are gone.
+// Answers how many it has. A user that cannot be made is said, the others
+// are made anyway.
+func (k *Keeper) users(ctx context.Context, want *llmgwv1.Slurm) (int32, error) {
+	made := k.madeUsers()
+	if !want.GetManageUsers() {
+		return 0, k.removeUsers(ctx, made, nil)
+	}
+	var errs []error
+	var keep []string
+	n := int32(0)
+	for _, u := range want.GetUsers() {
+		err := k.user(ctx, u, slices.Contains(made, u.GetLogin()), want.GetMakeHomes())
+		if errors.Is(err, errNotMade) {
+			errs = append(errs, err)
+			continue
+		}
+		keep = append(keep, u.GetLogin())
+		if err != nil {
+			errs = append(errs, err)
+			continue
+		}
+		n++
+	}
+	if err := k.keepMade(keep); err != nil {
+		return n, err
+	}
+	if err := k.removeUsers(ctx, made, keep); err != nil {
+		errs = append(errs, err)
+	}
+	if err := k.sshConf(ctx, len(keep) > 0); err != nil {
+		errs = append(errs, err)
+	}
+	// The controller is where they log in and submit from; a node only
+	// while a job of theirs runs there.
+	if err := k.nodeLogin(ctx, len(keep) > 0 && want.GetNode() && !want.GetController()); err != nil {
+		errs = append(errs, err)
+	}
+	return n, errors.Join(errs...)
+}
+
+// nodeLogin holds the users llmgw made here to their jobs when they log in
+// (pam_slurm_adopt), or no longer. Only they are: root and the site's
+// administrators log in as before, so a mistake here never locks out the
+// people who would put it right.
+func (k *Keeper) nodeLogin(ctx context.Context, on bool) error {
+	old, err := k.host.ReadFile(slurm.PAMSSHD)
+	if err != nil {
+		if !on {
+			return nil
+		}
+		return fmt.Errorf("reading %s: %w", slurm.PAMSSHD, err)
+	}
+	if on {
+		if err := k.adoptModule(ctx); err != nil {
+			return err
+		}
+	}
+	read := func(name string) (string, error) {
+		b, err := k.host.ReadFile(filepath.Join(filepath.Dir(slurm.PAMSSHD), name))
+		return string(b), err
+	}
+	stack, contained, err := slurm.PAMAdopt(string(old), k.usersMarker(), read, on)
+	if err != nil {
+		return err
+	}
+	if stack == string(old) {
+		return nil
+	}
+	if on {
+		k.log.Info("llmgw's Slurm users log in to this node only while a job of theirs runs here",
+			"in_the_job", contained)
+		if !contained {
+			k.log.Warn("pam_systemd could not be skipped for them safely here: their ssh sessions are not held " +
+				"to their job's cores and cards, only to having one")
+		}
+	} else {
+		k.log.Info("llmgw's Slurm users are no longer held to their jobs when they log in here")
+	}
+	// Read by sshd for each login: nothing to reload.
+	_, err = k.write(slurm.PAMSSHD, []byte(stack), 0o644, "")
+	return err
+}
+
+// adoptModule is there, or is installed.
+func (k *Keeper) adoptModule(ctx context.Context) error {
+	if k.installed[slurm.AdoptModule] {
+		return nil
+	}
+	find := func() bool {
+		for _, d := range slurm.ModuleDirs {
+			if _, err := k.host.Run(ctx, nil, "test", "-f", filepath.Join(d, slurm.AdoptModule)); err == nil {
+				return true
+			}
+		}
+		return false
+	}
+	if !find() {
+		if k.inst == nil {
+			return fmt.Errorf("%s missing, and this agent installs nothing", slurm.AdoptModule)
+		}
+		k.log.Info("installing pam_slurm_adopt, which holds llmgw's Slurm users to their jobs here")
+		if err := k.inst.InstallPackages(ctx, slurm.AdoptPackages); err != nil {
+			return err
+		}
+		if !find() {
+			return fmt.Errorf("%s is still missing after installing %s", slurm.AdoptModule,
+				strings.Join(slurm.AdoptPackages["deb"], ", "))
+		}
+	}
+	k.installed[slurm.AdoptModule] = true
+	return nil
+}
+
+// errNotMade is a user llmgw could not make here, nor took for its own.
+var errNotMade = errors.New("not made")
+
+type notMade struct{ msg string }
+
+func (e notMade) Error() string        { return e.msg }
+func (e notMade) Is(target error) bool { return target == errNotMade }
+
+// user keeps one user; ours is whether llmgw made it here before.
+func (k *Keeper) user(ctx context.Context, u *llmgwv1.SlurmUser, ours, makeHome bool) error {
+	login, uid, home := u.GetLogin(), u.GetUid(), u.GetHome()
+	if !slurm.ValidLogin(login) || uid < slurm.MinUID || !filepath.IsAbs(home) || strings.Contains(home, "..") {
+		return notMade{fmt.Sprintf("the leader sent a user %q, UID %d, home %q", login, uid, home)}
+	}
+	id := strconv.FormatUint(uint64(uid), 10)
+	gecos := strings.Map(func(r rune) rune {
+		if r == ':' || r == ',' || r == '\n' || r < ' ' {
+			return -1
+		}
+		return r
+	}, u.GetName())
+	byName, okName := k.passwd(ctx, login)
+	byID, okID := k.passwd(ctx, id)
+	switch {
+	case okName && byName[2] != id:
+		return notMade{fmt.Sprintf("%s is already a user here with UID %s, not llmgw's %s", login, byName[2], id)}
+	case !okName && okID:
+		return notMade{fmt.Sprintf("UID %s is already %s's here, so %s is not made", id, byID[0], login)}
+	case !okName:
+		if _, ok := k.group(ctx, login); !ok {
+			if out, err := k.host.Run(ctx, nil, "groupadd", "-g", id, login); err != nil {
+				return notMade{fmt.Sprintf("groupadd %s: %v: %s", login, err, lastLine(out))}
+			}
+		}
+		k.log.Info("a Slurm user is made here, as llmgw's accounts say", "login", login, "uid", uid)
+		if out, err := k.host.Run(ctx, nil, "useradd", "-u", id, "-g", id, "-M", "-d", home, "-s", "/bin/bash",
+			"-c", gecos, login); err != nil {
+			return notMade{fmt.Sprintf("useradd %s: %v: %s", login, err, lastLine(out))}
+		}
+	case ours && (byName[4] != gecos || byName[5] != home):
+		if out, err := k.host.Run(ctx, nil, "usermod", "-c", gecos, "-d", home, login); err != nil {
+			return fmt.Errorf("usermod %s: %v: %s", login, err, lastLine(out))
+		}
+	}
+	keys := strings.Join(u.GetSshKeys(), "\n")
+	file := filepath.Join(slurm.AuthorizedKeysDir, login)
+	if keys == "" {
+		_ = k.host.Remove(file)
+	} else if _, err := k.write(file, []byte(keys+"\n"), 0o644, ""); err != nil {
+		return err
+	}
+	if makeHome {
+		return k.home(ctx, login, id, home)
+	}
+	return nil
+}
+
+// home makes a user's home directory, from /etc/skel, when it is not there.
+func (k *Keeper) home(ctx context.Context, login, id, home string) error {
+	if _, err := k.host.Run(ctx, nil, "test", "-d", home); err == nil {
+		return nil
+	}
+	if err := k.host.MkdirAll(filepath.Dir(home), 0o755, ""); err != nil {
+		return err
+	}
+	k.log.Info("a Slurm user's home is made", "login", login, "home", home)
+	for _, c := range [][]string{
+		{"install", "-d", "-m", "0700", "-o", id, "-g", id, home},
+		{"cp", "-rT", "/etc/skel", home},
+		{"chown", "-R", id + ":" + id, home},
+	} {
+		if out, err := k.host.Run(ctx, nil, c[0], c[1:]...); err != nil {
+			return fmt.Errorf("making %s's home: %s: %v: %s", login, c[0], err, lastLine(out))
+		}
+	}
+	return nil
+}
+
+// removeUsers takes away the users llmgw made here that are not in keep.
+// Their homes and files stay; their UID is never given to another.
+func (k *Keeper) removeUsers(ctx context.Context, made, keep []string) error {
+	var errs []error
+	left := slices.Clone(keep)
+	for _, login := range made {
+		if slices.Contains(keep, login) {
+			continue
+		}
+		k.log.Info("a Slurm user llmgw made is taken away; its home stays", "login", login)
+		_ = k.host.Remove(filepath.Join(slurm.AuthorizedKeysDir, login))
+		if _, ok := k.passwd(ctx, login); ok {
+			if out, err := k.host.Run(ctx, nil, "userdel", login); err != nil {
+				// Still logged in, or running something: tried again.
+				errs = append(errs, fmt.Errorf("userdel %s: %v: %s", login, err, lastLine(out)))
+				left = append(left, login)
+				continue
+			}
+		}
+	}
+	if !slices.Equal(left, made) {
+		if err := k.keepMade(left); err != nil {
+			errs = append(errs, err)
+		}
+	}
+	if len(left) == 0 {
+		if err := k.sshConf(ctx, false); err != nil {
+			errs = append(errs, err)
+		}
+		if err := k.nodeLogin(ctx, false); err != nil {
+			errs = append(errs, err)
+		}
+	}
+	return errors.Join(errs...)
+}
+
+func (k *Keeper) madeUsers() []string {
+	b, err := k.host.ReadFile(k.usersMarker())
+	if err != nil {
+		return nil
+	}
+	return strings.Fields(string(b))
+}
+
+func (k *Keeper) keepMade(logins []string) error {
+	s := slices.Clone(logins)
+	slices.Sort(s)
+	s = slices.Compact(s)
+	if len(s) == 0 {
+		return k.host.Remove(k.usersMarker())
+	}
+	_, err := k.write(k.usersMarker(), []byte(strings.Join(s, "\n")+"\n"), 0o644, "")
+	return err
+}
+
+// sshConf has sshd read the users' keys from AuthorizedKeysDir, or no
+// longer.
+func (k *Keeper) sshConf(ctx context.Context, on bool) error {
+	if !on {
+		if _, err := k.host.ReadFile(slurm.SSHDropIn); err != nil {
+			return nil
+		}
+		_ = k.host.Remove(slurm.SSHDropIn)
+		return k.reloadSSH(ctx)
+	}
+	changed, err := k.write(slurm.SSHDropIn, []byte(slurm.SSHConf), 0o644, "")
+	if err != nil || !changed {
+		return err
+	}
+	if out, err := k.host.Run(ctx, nil, "sshd", "-t"); err != nil {
+		_ = k.host.Remove(slurm.SSHDropIn)
+		return fmt.Errorf("sshd refuses llmgw's setting for the users' keys, so it is not kept: %v: %s", err, lastLine(out))
+	}
+	k.log.Info("sshd reads the Slurm users' keys from llmgw too", "dir", slurm.AuthorizedKeysDir)
+	return k.reloadSSH(ctx)
+}
+
+// reloadSSH has sshd read its settings again, under Debian's name for it
+// or Red Hat's; one that is not running reads them when it starts.
+func (k *Keeper) reloadSSH(ctx context.Context) error {
+	for _, u := range []string{"ssh", "sshd"} {
+		if k.active(ctx, u) {
+			return k.systemctl(ctx, "reload", u)
+		}
+	}
+	return nil
+}
+
+// passwd is getent's line for a user, by login or UID, in its fields.
+func (k *Keeper) passwd(ctx context.Context, key string) ([]string, bool) {
+	out, err := k.host.Run(ctx, nil, "getent", "passwd", key)
+	f := strings.Split(strings.TrimSpace(string(out)), ":")
+	if err != nil || len(f) < 7 {
+		return nil, false
+	}
+	return f, true
+}
+
+func (k *Keeper) group(ctx context.Context, key string) ([]string, bool) {
+	out, err := k.host.Run(ctx, nil, "getent", "group", key)
+	f := strings.Split(strings.TrimSpace(string(out)), ":")
+	if err != nil || len(f) < 3 {
+		return nil, false
+	}
+	return f, true
+}

internal/agent/supervisor/health.go

@@ -54,9 +54,13 @@ func NewHTTPHealth(timeout time.Duration, opts ...HTTPHealthOption) *HTTPHealth
 
 // Ready reports whether the engine at baseURL is serving model.
 func (h *HTTPHealth) Ready(ctx context.Context, baseURL, model, apiKey string) error {
-	if err := h.get(ctx, baseURL+h.healthPath, apiKey, nil); err != nil {
+	var body healthBody
+	if err := h.get(ctx, baseURL+h.healthPath, apiKey, &body); err != nil {
 		return fmt.Errorf("health: %w", err)
 	}
+	if body.notReady() {
+		return fmt.Errorf("health: the engine says %q", body.Status)
+	}
 	if !h.requireName || model == "" {
 		return nil
 	}
@@ -98,5 +102,29 @@ func (h *HTTPHealth) get(ctx context.Context, url, apiKey string, out any) error
 	if out == nil {
 		return nil
 	}
+	if hb, ok := out.(*healthBody); ok {
+		// Most engines answer /health with nothing, or with text: only a
+		// JSON status says anything, and a body that is not one is not a
+		// failure.
+		_ = json.NewDecoder(io.LimitReader(resp.Body, 64<<10)).Decode(hb)
+		return nil
+	}
 	return json.NewDecoder(resp.Body).Decode(out)
 }
+
+// healthBody is what an engine's /health may say in its body. Some answer
+// 200 from the moment their HTTP server is up and say in the body that the
+// model is still loading — FreeToken's {"status":"loading","phase":...} —
+// and a model that lists itself in /v1/models meanwhile would be routed to.
+type healthBody struct {
+	Status string `json:"status"`
+}
+
+// notReady is a body that says the engine cannot serve yet, or has failed.
+func (b healthBody) notReady() bool {
+	switch b.Status {
+	case "loading", "starting", "initializing", "warming_up", "error", "unhealthy":
+		return true
+	}
+	return false
+}

internal/agent/supervisor/health_test.go

@@ -103,3 +103,29 @@ func TestTheHealthCheckSendsTheEnginesKey(t *testing.T) {
 		t.Error("without the key the engine refused, and that read as ready")
 	}
 }
+
+func TestNotReadyWhileTheHealthBodySaysLoading(t *testing.T) {
+	// FreeToken answers 200 from the moment its HTTP server is up and lists
+	// its model meanwhile; only the body says the weights are still loading.
+	body := `{"status":"loading","phase":"weights","progress":{"done_bytes":1,"total_bytes":9},"model":"glm"}`
+	mux := http.NewServeMux()
+	mux.HandleFunc("GET /health", func(w http.ResponseWriter, _ *http.Request) { io.WriteString(w, body) })
+	mux.HandleFunc("GET /v1/models", func(w http.ResponseWriter, _ *http.Request) {
+		io.WriteString(w, `{"data":[{"id":"glm"}]}`)
+	})
+	srv := httptest.NewServer(mux)
+	t.Cleanup(srv.Close)
+	h := NewHTTPHealth(time.Second)
+	if err := h.Ready(t.Context(), srv.URL, "glm", ""); err == nil || !strings.Contains(err.Error(), "loading") {
+		t.Errorf("a loading engine was reported ready, or not said to be loading: %v", err)
+	}
+	body = `{"status":"ok","model":"glm","uptime_s":3}`
+	if err := h.Ready(t.Context(), srv.URL, "glm", ""); err != nil {
+		t.Errorf("a serving engine was reported unready: %v", err)
+	}
+	// Text, as some engines answer: not a status, and not a failure.
+	body = "OK"
+	if err := h.Ready(t.Context(), srv.URL, "glm", ""); err != nil {
+		t.Errorf("a plain-text health answer was taken for a failure: %v", err)
+	}
+}

internal/agent/supervisor/process_unix.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"os"
 	"strconv"
+	"strings"
 	"syscall"
 )
 
@@ -63,3 +64,71 @@ func isZombie(pid int) bool {
 	}
 	return b[i+2] == 'Z'
 }
+
+// strays are a process's descendants that left its group, as an engine that
+// starts its worker in a session of its own does (Strata): signalling the
+// group misses them, and one left behind holds the card and the RAM the next
+// replica needs. Found before the group is signalled, while they are still
+// its descendants: after, they belong to init. Linux only; elsewhere none.
+func strays(pid int) []int {
+	entries, err := os.ReadDir("/proc")
+	if err != nil {
+		return nil
+	}
+	parent := map[int]int{}
+	group := map[int]int{}
+	for _, e := range entries {
+		p, err := strconv.Atoi(e.Name())
+		if err != nil {
+			continue
+		}
+		if pp, pg, ok := parentAndGroup(p); ok {
+			parent[p], group[p] = pp, pg
+		}
+	}
+	var out []int
+	for p := range parent {
+		if group[p] == pid || p == pid {
+			continue
+		}
+		// Bounded: no engine's tree is this deep, and a pid reused while
+		// /proc was read must not loop.
+		for q, n := parent[p], 0; q > 1 && n < 32; q, n = parent[q], n+1 {
+			if q == pid {
+				out = append(out, p)
+				break
+			}
+		}
+	}
+	return out
+}
+
+// parentAndGroup reads a process's parent and group from /proc.
+func parentAndGroup(pid int) (int, int, bool) {
+	b, err := os.ReadFile("/proc/" + strconv.Itoa(pid) + "/stat")
+	if err != nil {
+		return 0, 0, false
+	}
+	i := bytes.LastIndexByte(b, ')')
+	if i < 0 {
+		return 0, 0, false
+	}
+	f := strings.Fields(string(b[i+1:]))
+	if len(f) < 3 {
+		return 0, 0, false
+	}
+	pp, err1 := strconv.Atoi(f[1])
+	pg, err2 := strconv.Atoi(f[2])
+	return pp, pg, err1 == nil && err2 == nil
+}
+
+// signalStrays signals each stray, and its own group when it leads one.
+func signalStrays(pids []int, sig syscall.Signal) {
+	for _, p := range pids {
+		if _, pg, ok := parentAndGroup(p); ok && pg == p {
+			_ = syscall.Kill(-p, sig)
+			continue
+		}
+		_ = syscall.Kill(p, sig)
+	}
+}

internal/agent/supervisor/supervisor.go

@@ -582,10 +582,24 @@ func (s *Supervisor) Stop(ctx context.Context, id string) (err error) {
 	}
 
 	// Signal the group, not the process: engines fork workers that hold GPU
-	// memory, and those outlive a signal sent only to the leader.
+	// memory, and those outlive a signal sent only to the leader. And what
+	// left the group, found while it is still the leader's.
+	left := strays(pid)
 	if err := killGroup(pid, syscall.SIGTERM); err != nil && !errors.Is(err, os.ErrProcessDone) {
 		s.log.Warn("SIGTERM failed", "shard", id, "err", err)
 	}
+	signalStrays(left, syscall.SIGTERM)
+	defer func() {
+		// Given as long as the group to go; then killed.
+		deadline := time.Now().Add(5 * time.Second)
+		for time.Now().Before(deadline) && slices.ContainsFunc(left, processAlive) {
+			time.Sleep(100 * time.Millisecond)
+		}
+		if slices.ContainsFunc(left, processAlive) {
+			s.log.Warn("killing an engine's processes that left its group", "shard", id, "pids", left)
+			signalStrays(left, syscall.SIGKILL)
+		}
+	}()
 
 	// force kills the group outright. exec.Cmd's cancel does this for shards we
 	// started; an adopted shard has none, so signal its group directly.
@@ -598,6 +612,7 @@ func (s *Supervisor) Stop(ctx context.Context, id string) (err error) {
 		if err := killGroup(pid, syscall.SIGKILL); err != nil && !errors.Is(err, os.ErrProcessDone) {
 			s.log.Warn("SIGKILL failed", "shard", id, "err", err)
 		}
+		signalStrays(left, syscall.SIGKILL)
 	}
 
 	select {

internal/agent/supervisor/supervisor_test.go

@@ -8,6 +8,7 @@ import (
 	"log/slog"
 	"net"
 	"os"
+	"os/exec"
 	"path/filepath"
 	"slices"
 	"strings"
@@ -981,3 +982,38 @@ func TestAnOutputFileIsEmptiedPastItsBound(t *testing.T) {
 		t.Errorf("a line written after the file was emptied was lost: %v", sh.Logs(8))
 	}
 }
+
+// An engine's worker in a session of its own, as Strata's is, is not in the
+// group a stop signals: it is found and stopped as well.
+func TestStopKillsAWorkerThatLeftTheGroup(t *testing.T) {
+	if _, err := os.Stat("/proc/self/stat"); err != nil {
+		t.Skip("needs /proc")
+	}
+	if _, err := exec.LookPath("setsid"); err != nil {
+		t.Skip("needs setsid")
+	}
+	s, _ := newSupervisor(t)
+	marker := filepath.Join(t.TempDir(), "worker.pid")
+	script := fmt.Sprintf(`setsid sh -c 'echo $$ > %s; sleep 30' & sleep 30`, marker)
+	if _, err := s.Start(t.Context(), "a", engines.ShardSpec{}, shellCommand(script)); err != nil {
+		t.Fatal(err)
+	}
+	var worker int
+	for deadline := time.Now().Add(3 * time.Second); time.Now().Before(deadline) && worker == 0; time.Sleep(10 * time.Millisecond) {
+		if b, err := os.ReadFile(marker); err == nil {
+			_, _ = fmt.Sscanf(strings.TrimSpace(string(b)), "%d", &worker)
+		}
+	}
+	if worker == 0 {
+		t.Fatal("the worker never started")
+	}
+	if err := s.Stop(t.Context(), "a"); err != nil {
+		t.Fatal(err)
+	}
+	for deadline := time.Now().Add(8 * time.Second); time.Now().Before(deadline); time.Sleep(50 * time.Millisecond) {
+		if !processAlive(worker) {
+			return
+		}
+	}
+	t.Errorf("the worker %d that left the group survived the stop", worker)
+}

internal/agent/upgrade.go

@@ -35,12 +35,25 @@ func (a *Agent) upgrade(ctx context.Context, m *llmgwv1.UpgradeAgent, out chan<-
 		select {
 		case out <- &llmgwv1.AgentMessage{Msg: &llmgwv1.AgentMessage_Upgrade{Upgrade: reply}}:
 		case <-ctx.Done():
+		case <-time.After(5 * time.Second):
 		}
 	}
+	// One at a time: a second request while one downloads would race it to
+	// the same file.
+	if !a.upgrading.CompareAndSwap(false, true) {
+		reply.Error = "an upgrade is already under way on this node"
+		send()
+		return
+	}
+	defer a.upgrading.Store(false)
 	a.log.Info("upgrading", "to", m.GetVersion(), "from", a.cfg.Version)
 	self, err := upgradeTarget()
 	if err == nil {
-		dctx, cancel := context.WithTimeout(ctx, 10*time.Minute)
+		// Not the stream's context: a stream that breaks while the release
+		// downloads — a control plane restarting, leadership moving — cancelled
+		// the upgrade with nobody told. The node registering again on the
+		// version is what the control plane watches for.
+		dctx, cancel := context.WithTimeout(context.WithoutCancel(ctx), upgradeDownload)
 		_, err = release.Install(dctx, nil, m.GetReleases(), m.GetVersion(), runtime.GOOS, runtime.GOARCH, self)
 		cancel()
 	}
@@ -53,6 +66,13 @@ func (a *Agent) upgrade(ctx context.Context, m *llmgwv1.UpgradeAgent, out chan<-
 	send()
 	a.log.Info("upgraded; restarting into the new version", "to", m.GetVersion())
 	// A moment for the reply to leave before the stream goes.
-	time.Sleep(500 * time.Millisecond)
+	time.Sleep(upgradeLinger)
 	exitForUpgrade()
 }
+
+// upgradeDownload bounds an upgrade's download; the control plane waits
+// longer (agentsvc.UpgradeWait). upgradeLinger is the moment its answer has
+// to leave before the agent exits; a variable so tests need not wait it.
+const upgradeDownload = 10 * time.Minute
+
+var upgradeLinger = 2 * time.Second

internal/agent/weights/fetch.go

@@ -38,8 +38,14 @@ type Progress struct {
 	Source     string
 	BytesDone  int64
 	BytesTotal int64
+	// Phase is empty while the bytes arrive, PhaseChecking while a set's
+	// files are checked against the mirror's checksums.
+	Phase string
 }
 
+// PhaseChecking is a set's files being checked, after they arrived.
+const PhaseChecking = "checking"
+
 // Percent is how far along, or -1 when the total is unknown.
 func (p Progress) Percent() int {
 	if p.BytesTotal <= 0 {
@@ -90,14 +96,27 @@ var ErrNotLocal = errors.New("the weights are not on this node and the source sa
 // already cached, downloaded, or downloaded by another shard this one waited
 // for — since that is most of the difference between a replica that starts
 // in a minute and one that takes an hour.
-func (f *Fetcher) Ensure(ctx context.Context, source string) (path string, err error) {
+//
+// A source in a pickle format, or a set with such a file in it, is refused
+// before anything is fetched unless allowPickle; see Pickled.
+func (f *Fetcher) Ensure(ctx context.Context, source string, allowPickle bool) (path string, err error) {
+	return f.EnsureChecked(ctx, source, allowPickle, "")
+}
+
+// EnsureChecked is Ensure for a file whose sha256 the registry published: a
+// copy that is not it, downloaded or cached, is refused rather than served.
+// An empty sha256 checks nothing.
+func (f *Fetcher) EnsureChecked(ctx context.Context, source string, allowPickle bool, sha256 string) (path string, err error) {
 	ctx, span := obs.Tracer(tracerName).Start(ctx, "fetch weights",
 		trace.WithAttributes(attribute.String("llmgw.weights.source", redact(source))))
 	defer func() {
 		span.SetAttributes(attribute.String("llmgw.weights.path", path))
 		obs.End(span, err)
 	}()
-	return f.ensure(ctx, source)
+	if !allowPickle && !IsSet(source) && Pickled(source) {
+		return "", refusePickle(source)
+	}
+	return f.ensure(ctx, source, allowPickle, strings.ToLower(sha256))
 }
 
 // tracerName names the spans this package opens.
@@ -119,7 +138,7 @@ func redact(source string) string {
 	return u.String()
 }
 
-func (f *Fetcher) ensure(ctx context.Context, source string) (string, error) {
+func (f *Fetcher) ensure(ctx context.Context, source string, allowPickle bool, want string) (string, error) {
 	if source == "" {
 		return "", fmt.Errorf("no source given")
 	}
@@ -127,7 +146,11 @@ func (f *Fetcher) ensure(ctx context.Context, source string) (string, error) {
 	// file, and a set has to be handled before any of it.
 	if IsSet(source) {
 		outcome(ctx, "set")
-		return f.ensureSet(ctx, source)
+		// Shared like a single file: four replicas placed on one node's
+		// four cards want the same set at the same instant, and each
+		// fetching it wrote the same .part files over one another — three
+		// failed renaming a file another had already renamed.
+		return f.shared(ctx, source, func() (string, error) { return f.ensureSet(ctx, source, allowPickle) })
 	}
 	if !isRemote(source) {
 		if !isPath(source) {
@@ -150,10 +173,24 @@ func (f *Fetcher) ensure(ctx context.Context, source string) (string, error) {
 		return "", err
 	}
 	if _, err := os.Stat(target); err == nil {
-		outcome(ctx, "cached")
-		return target, nil
+		if ok, err := f.cachedIs(target, want); err != nil {
+			return "", err
+		} else if ok {
+			outcome(ctx, "cached")
+			return target, nil
+		}
+		// Not the file the catalog names: fetched again below, and checked.
 	}
 
+	return f.shared(ctx, source, func() (string, error) {
+		outcome(ctx, "downloaded")
+		return f.fetchChecked(ctx, source, target, want)
+	})
+}
+
+// shared runs fetch once for every caller asking for source at the same
+// time; the others wait for its answer.
+func (f *Fetcher) shared(ctx context.Context, source string, fetch func() (string, error)) (string, error) {
 	f.mu.Lock()
 	if f.inflight == nil {
 		f.inflight = map[string]*fetchCall{}
@@ -172,8 +209,7 @@ func (f *Fetcher) ensure(ctx context.Context, source string) (string, error) {
 	f.inflight[source] = call
 	f.mu.Unlock()
 
-	outcome(ctx, "downloaded")
-	call.path, call.err = f.fetch(ctx, source, target)
+	call.path, call.err = fetch()
 	close(call.done)
 
 	f.mu.Lock()
@@ -184,9 +220,83 @@ func (f *Fetcher) ensure(ctx context.Context, source string) (string, error) {
 
 // fetch downloads a source to a target, resuming and verifying.
 func (f *Fetcher) fetch(ctx context.Context, source, target string) (string, error) {
-	return f.fetchAs(ctx, source, source, target)
+	return f.fetchAs(ctx, source, source, target, setShare{})
 }
 
+// fetchChecked is fetch for a file whose sha256 is known: a download that
+// is not it is deleted and refused, naming both, and one that is has its
+// checksum kept beside it so a cached copy is not read again at each start.
+func (f *Fetcher) fetchChecked(ctx context.Context, source, target, want string) (string, error) {
+	got, err := f.fetch(ctx, source, target)
+	if err != nil || want == "" {
+		return got, err
+	}
+	sum, err := fileSHA256(target)
+	if err != nil {
+		return "", err
+	}
+	if sum != want {
+		_ = os.Remove(target)
+		return "", fmt.Errorf("%s arrived with sha256 %s, and the registry published %s for it: refused and deleted",
+			redact(source), sum, want)
+	}
+	_ = writeChecksum(target, sum)
+	return target, nil
+}
+
+// cachedIs says whether a cached file is the one whose sha256 is want: by
+// the checksum kept beside it, or worked out once and then kept. One that
+// is not is deleted, to be fetched again. Any file is, when want is empty.
+func (f *Fetcher) cachedIs(target, want string) (bool, error) {
+	kept := readChecksum(target)
+	if want == "" || kept == want {
+		return true, nil
+	}
+	if kept != "" {
+		// Checked when it was fetched, as another file: the same address
+		// gives the same bytes again, and another model may be serving
+		// these. Refused, and left.
+		return false, fmt.Errorf("the copy of this file here has sha256 %s, as fetched, and the catalog names %s: refused",
+			kept, want)
+	}
+	sum, err := fileSHA256(target)
+	if err != nil {
+		return false, err
+	}
+	if sum == want {
+		_ = writeChecksum(target, sum)
+		return true, nil
+	}
+	if f.Log != nil {
+		f.Log.Warn("a cached file is not the one the catalog names; fetching it again",
+			"path", target, "sha256", sum, "want", want)
+	}
+	_ = os.Remove(checksumPath(target))
+	return false, os.Remove(target)
+}
+
+// checksumPath is where a cached file's sha256 is kept: beside it, hidden,
+// so no scan takes it for weights.
+func checksumPath(file string) string {
+	return filepath.Join(filepath.Dir(file), "."+filepath.Base(file)+".sha256")
+}
+
+func writeChecksum(file, sum string) error {
+	return os.WriteFile(checksumPath(file), []byte(sum+"\n"), 0o644) //nolint:gosec // a checksum, for anyone to read
+}
+
+func readChecksum(file string) string {
+	b, err := os.ReadFile(checksumPath(file)) //nolint:gosec // beside a file this fetcher wrote
+	if err != nil {
+		return ""
+	}
+	return strings.TrimSpace(string(b))
+}
+
+// ChecksumPath is where the sha256 of a cached file is kept, for whoever
+// deletes the file to delete with it.
+func ChecksumPath(file string) string { return checksumPath(file) }
+
 // fetchAs downloads one URL but reports progress under another name.
 //
 // The two differ for a set: the shards waiting on it asked for the version, not
@@ -195,7 +305,7 @@ func (f *Fetcher) fetch(ctx context.Context, source, target string) (string, err
 //
 // Each file is a "download" span, with how many bytes it took and where a
 // resumed one started from.
-func (f *Fetcher) fetchAs(ctx context.Context, source, reportAs, target string) (_ string, err error) {
+func (f *Fetcher) fetchAs(ctx context.Context, source, reportAs, target string, share setShare) (_ string, err error) {
 	ctx, span := obs.Tracer(tracerName).Start(ctx, "download",
 		trace.WithSpanKind(trace.SpanKindClient),
 		trace.WithAttributes(attribute.String("llmgw.weights.source", redact(source))))
@@ -259,8 +369,8 @@ func (f *Fetcher) fetchAs(ctx context.Context, source, reportAs, target string)
 	}
 
 	total := existing + resp.ContentLength
-	f.report(Progress{Source: reportAs, BytesDone: existing, BytesTotal: total})
-	written, err := f.copyWithProgress(ctx, out, resp.Body, reportAs, existing, total)
+	f.report(share.of(Progress{Source: reportAs, BytesDone: existing, BytesTotal: total}))
+	written, err := f.copyWithProgress(ctx, out, resp.Body, reportAs, existing, total, share)
 	span.SetAttributes(
 		attribute.Int64("llmgw.weights.resumed_from", existing),
 		attribute.Int64("llmgw.weights.bytes", written),
@@ -284,7 +394,7 @@ func (f *Fetcher) fetchAs(ctx context.Context, source, reportAs, target string)
 }
 
 func (f *Fetcher) copyWithProgress(ctx context.Context, dst io.Writer, src io.Reader,
-	source string, done, total int64,
+	source string, done, total int64, share setShare,
 ) (int64, error) {
 	buf := make([]byte, 4<<20)
 	last := time.Now()
@@ -302,12 +412,12 @@ func (f *Fetcher) copyWithProgress(ctx context.Context, dst io.Writer, src io.Re
 			// hundred gigabytes is twenty-five thousand callbacks, and nobody
 			// needs to be told that often.
 			if time.Since(last) > 2*time.Second {
-				f.report(Progress{Source: source, BytesDone: done, BytesTotal: total})
+				f.report(share.of(Progress{Source: source, BytesDone: done, BytesTotal: total}))
 				last = time.Now()
 			}
 		}
 		if errors.Is(err, io.EOF) {
-			f.report(Progress{Source: source, BytesDone: done, BytesTotal: total})
+			f.report(share.of(Progress{Source: source, BytesDone: done, BytesTotal: total}))
 			return done, nil
 		}
 		if err != nil {
@@ -316,6 +426,18 @@ func (f *Fetcher) copyWithProgress(ctx context.Context, dst io.Writer, src io.Re
 	}
 }
 
+// setShare places one file's download within its set's: the bytes of the
+// files before it, and the whole set's. Zero for a file on its own.
+type setShare struct{ before, total int64 }
+
+// of is a file's progress as the set's.
+func (s setShare) of(p Progress) Progress {
+	if s.total > 0 {
+		p.BytesDone, p.BytesTotal = s.before+p.BytesDone, s.total
+	}
+	return p
+}
+
 func (f *Fetcher) report(p Progress) {
 	if f.OnProgress != nil {
 		f.OnProgress(p)

internal/agent/weights/fetch_test.go

@@ -1,6 +1,8 @@
 package weights
 
 import (
+	"crypto/sha256"
+	"encoding/hex"
 	"errors"
 	"net/http"
 	"net/http/httptest"
@@ -37,7 +39,7 @@ func TestALocalPathIsUsedNeverCopied(t *testing.T) {
 	}
 	f := &Fetcher{CacheDir: t.TempDir()}
 
-	got, err := f.Ensure(t.Context(), path)
+	got, err := f.Ensure(t.Context(), path, false)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -53,7 +55,7 @@ func TestAMissingLocalPathSaysWhatIsWrong(t *testing.T) {
 	// The commonest setup mistake, and "no such file" from an engine twenty
 	// minutes into a load is a poor way to learn it.
 	f := &Fetcher{CacheDir: t.TempDir()}
-	_, err := f.Ensure(t.Context(), filepath.Join(t.TempDir(), "absent.gguf"))
+	_, err := f.Ensure(t.Context(), filepath.Join(t.TempDir(), "absent.gguf"), false)
 	if !errors.Is(err, ErrNotLocal) {
 		t.Fatalf("err = %v, want ErrNotLocal", err)
 	}
@@ -67,7 +69,7 @@ func TestARemoteFileIsFetchedOnceAndReused(t *testing.T) {
 	srv := serveBytes(t, []byte("the weights"), &hits)
 	f := &Fetcher{CacheDir: t.TempDir()}
 
-	first, err := f.Ensure(t.Context(), srv.URL+"/model.gguf")
+	first, err := f.Ensure(t.Context(), srv.URL+"/model.gguf", false)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -76,7 +78,7 @@ func TestARemoteFileIsFetchedOnceAndReused(t *testing.T) {
 		t.Fatalf("fetched content = %q, err %v", body, err)
 	}
 
-	second, err := f.Ensure(t.Context(), srv.URL+"/model.gguf")
+	second, err := f.Ensure(t.Context(), srv.URL+"/model.gguf", false)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -103,7 +105,7 @@ func TestConcurrentRequestsProduceOneDownload(t *testing.T) {
 		wg.Add(1)
 		go func() {
 			defer wg.Done()
-			paths[i], errs[i] = f.Ensure(t.Context(), srv.URL+"/model.gguf")
+			paths[i], errs[i] = f.Ensure(t.Context(), srv.URL+"/model.gguf", false)
 		}()
 	}
 	wg.Wait()
@@ -139,7 +141,7 @@ func TestAPartialDownloadIsNeverVisibleAsTheModel(t *testing.T) {
 
 	cache := t.TempDir()
 	f := &Fetcher{CacheDir: cache}
-	if _, err := f.Ensure(t.Context(), srv.URL+"/model.gguf"); err == nil {
+	if _, err := f.Ensure(t.Context(), srv.URL+"/model.gguf", false); err == nil {
 		t.Fatal("a truncated download was reported as success")
 	}
 	entries, _ := os.ReadDir(cache)
@@ -169,7 +171,7 @@ func TestAnInterruptedDownloadResumes(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	got, err := f.Ensure(t.Context(), srv.URL+"/model.gguf")
+	got, err := f.Ensure(t.Context(), srv.URL+"/model.gguf", false)
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -232,7 +234,7 @@ func TestProgressIsReportedWithAPercentage(t *testing.T) {
 		CacheDir:   t.TempDir(),
 		OnProgress: func(p Progress) { seen = append(seen, p) },
 	}
-	if _, err := f.Ensure(t.Context(), srv.URL+"/model.gguf"); err != nil {
+	if _, err := f.Ensure(t.Context(), srv.URL+"/model.gguf", false); err != nil {
 		t.Fatal(err)
 	}
 	if len(seen) == 0 {
@@ -254,7 +256,7 @@ func TestPercentIsUnknownRatherThanZeroWithoutATotal(t *testing.T) {
 
 func TestNoCacheDirectoryIsAClearRefusal(t *testing.T) {
 	f := &Fetcher{}
-	_, err := f.Ensure(t.Context(), "https://example.invalid/model.gguf")
+	_, err := f.Ensure(t.Context(), "https://example.invalid/model.gguf", false)
 	if err == nil || !strings.Contains(err.Error(), "cache directory") {
 		t.Errorf("err = %v, want a complaint about the cache directory", err)
 	}
@@ -269,7 +271,7 @@ func TestABareModelIdIsLeftToTheEngine(t *testing.T) {
 		"Qwen/Qwen2.5-7B-Instruct",
 		"meta-llama/Llama-3.3-70B-Instruct",
 	} {
-		got, err := f.Ensure(t.Context(), id)
+		got, err := f.Ensure(t.Context(), id, false)
 		if err != nil {
 			t.Errorf("%s was refused: %v", id, err)
 		}
@@ -283,7 +285,79 @@ func TestAnAbsolutePathThatIsMissingIsStillAnError(t *testing.T) {
 	// The other half: an operator writing a path means a file, and a missing
 	// one should be said plainly rather than handed to an engine.
 	f := &Fetcher{CacheDir: t.TempDir()}
-	if _, err := f.Ensure(t.Context(), "/models/definitely-not-here.gguf"); !errors.Is(err, ErrNotLocal) {
+	if _, err := f.Ensure(t.Context(), "/models/definitely-not-here.gguf", false); !errors.Is(err, ErrNotLocal) {
 		t.Errorf("err = %v, want ErrNotLocal", err)
 	}
 }
+
+// A file whose sha256 the registry published is served only as that file:
+// changed upstream after the model was added, it is refused with both
+// checksums named, and a cached copy that is not it is fetched again.
+func TestAFileThatIsNotWhatTheRegistryPublishedIsRefused(t *testing.T) {
+	body := "the weights as published"
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		_, _ = w.Write([]byte(body))
+	}))
+	defer srv.Close()
+	sum := sha256.Sum256([]byte(body))
+	want := hex.EncodeToString(sum[:])
+	f := &Fetcher{CacheDir: t.TempDir()}
+	src := srv.URL + "/model.gguf"
+
+	path, err := f.EnsureChecked(t.Context(), src, false, strings.ToUpper(want))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if readChecksum(path) != want {
+		t.Errorf("the checksum was not kept beside it: %q", readChecksum(path))
+	}
+
+	// The cached copy changed under it: fetched again, and the right one kept.
+	if err := os.WriteFile(path, []byte("tampered"), 0o644); err != nil {
+		t.Fatal(err)
+	}
+	_ = os.Remove(checksumPath(path))
+	if _, err := f.EnsureChecked(t.Context(), src, false, want); err != nil {
+		t.Fatal(err)
+	}
+	if b, _ := os.ReadFile(path); string(b) != body {
+		t.Errorf("the cached copy that was not it was served: %q", b)
+	}
+
+	// Upstream changed after the model was added.
+	body = "weights someone replaced"
+	f2 := &Fetcher{CacheDir: t.TempDir()}
+	_, err = f2.EnsureChecked(t.Context(), src, false, want)
+	if err == nil || !strings.Contains(err.Error(), want) || !strings.Contains(err.Error(), "sha256") {
+		t.Fatalf("not refused, or without both checksums: %v", err)
+	}
+	if _, err := os.Stat(CachePathFor(f2.CacheDir, src)); !os.IsNotExist(err) {
+		t.Error("the file that was not it was kept")
+	}
+	// With nothing to check against, it is fetched as before.
+	if _, err := f2.Ensure(t.Context(), src, false); err != nil {
+		t.Error(err)
+	}
+}
+
+// A cached file checked when it was fetched, that another model names with
+// another checksum, is refused and left for the model it is.
+func TestACheckedFileAnotherChecksumNamesIsRefusedAndLeft(t *testing.T) {
+	body := "the weights"
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) { _, _ = w.Write([]byte(body)) }))
+	defer srv.Close()
+	sum := sha256.Sum256([]byte(body))
+	f := &Fetcher{CacheDir: t.TempDir()}
+	src := srv.URL + "/m.gguf"
+	path, err := f.EnsureChecked(t.Context(), src, false, hex.EncodeToString(sum[:]))
+	if err != nil {
+		t.Fatal(err)
+	}
+	wrong := strings.Repeat("0", 64)
+	if _, err := f.EnsureChecked(t.Context(), src, false, wrong); err == nil || !strings.Contains(err.Error(), wrong) {
+		t.Fatalf("not refused: %v", err)
+	}
+	if b, err := os.ReadFile(path); err != nil || string(b) != body {
+		t.Errorf("the checked file was not left: %v", err)
+	}
+}

internal/agent/weights/gc.go

@@ -33,6 +33,17 @@ func CachePathFor(dir, source string) string {
 	return filepath.Join(dir, hex.EncodeToString(sum[:8])+"-"+name)
 }
 
+// cacheSource is the source a cache name is computed from: a set's address
+// without its trailing slash, as its directory was named when it was fetched.
+// Computed from the slash-ended address, every set in the catalog looked like
+// nobody's, and a sweep offered 330 GB of models still in use for deletion.
+func cacheSource(source string) string {
+	if IsSet(source) {
+		return strings.TrimSuffix(source, "/")
+	}
+	return source
+}
+
 // Owned reports whether a path is one this package created.
 //
 // The dividing line of the whole feature. A file without the cache name was
@@ -65,7 +76,7 @@ func Orphans(have []string, wanted []string) []string {
 		// By base name rather than full path: the control plane does not know
 		// each node's cache directory, and the name is what identifies the
 		// file wherever the node keeps it.
-		keep[filepath.Base(CachePathFor("", src))] = struct{}{}
+		keep[filepath.Base(CachePathFor("", cacheSource(src)))] = struct{}{}
 	}
 
 	var out []string