Compare commits

...
15 Commits
Author SHA1 Message Date
Ilker ManapandClaude Opus 5.5 2005cea326 v0.14.8: changelog and README
CI / test (push) Successful in 9m12s
Release / build (push) Successful in 4m20s
Release / notes (push) Successful in 3s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-18 11:00:28 +02:00
Ilker ManapandClaude Opus 5.5 f27d722111 People sign in with their Active Directory account
auth.ldap checks a username and password against the domain controllers
over LDAPS or StartTLS: found with a service account, checked by binding
as the person. Groups, nested ones included, give staff roles and places
in the tree. Accounts are tied to the objectGUID; the leader reads the
directory every 15 minutes, suspends whom it disabled or removed and
reinstates them after, and refuses a sync that would suspend most.

Entra ID sends no groups for someone in more than 200, only where to
fetch them; that was read as none and took an administrator's role away.
It is now left as it was, and logged. An account made by SCIM or OIDC
can be reinstated again.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-18 10:59:50 +02:00
Ilker ManapandClaude Opus 5.5 fe310576e7 Several keys for one outside provider
CI / test (push) Successful in 8m52s
An external model takes api_key_files, each a file and an optional weight,
in place of api_key_file. Each key is an endpoint of its own
(external:MODEL#1, #2), so the router's scoring, retries and breakers work
per key. A key answering 429 is passed over while its breaker is open. A
key answering 401 is passed over too, and external_key_refused is sent,
naming the endpoint and never the key; with a single key the 401 is still
the client's. A key whose file cannot be read is left out, and usage
records which key answered by its endpoint.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-17 08:56:30 +02:00
Ilker ManapandClaude Opus 5.5 688b205604 What llmgw adds to a request, measured: ctl bench overhead
CI / test (push) Successful in 8m56s
An engine that answers at once and the gateway in front of it, in one
process on loopback, given the same load straight and through the gateway:
bare, with authentication and accounting, and with the guardrails' pattern
checks. It reports p50 and p99, what the gateway added, and a rough CPU
per thousand requests.

On an Apple M5 at 100 to 2,000 requests a second (10 s a step, 72,000
requests, none failed), the gateway added 0.05–0.42 ms at p50 and at most
0.49 ms at p99. That is now in the README, with how it was measured and
why the CPU figure is rough. CI runs a one-second version.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-17 00:23:26 +02:00
Ilker ManapandClaude Opus 5.5 35ccabac28 modelserve: a job's request is set under the lock Get reads it with
CI / test (push) Successful in 8m49s
CI run 394, on a commit that changed only BACKLOG.md, failed with a data
race in TestAVisionModelIsServedWithItsProjectorBesideIt. The serve job's
goroutine wrote j.Request without the lock (since efffdb6, which sized
embedding slots), while Get's snapshot read the job under it. Every other
write to a running job holds the lock. It did not reproduce in 30 runs
here; the race detector's trace names this write and that read.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 22:40:09 +02:00
Ilker ManapandClaude Opus 5.5 2545fca174 An upgrade refits the judge an earlier server init installed
CI / test (push) Successful in 8m53s
A v0.14.6 install on fewer than eight cores refused every request until
its judge was given cores by hand, and upgrading did not change that. The
judge keeper, on the leader, now finds a judge whose entry is exactly what
init before v0.14.7 wrote: one llama.cpp CPU profile, an 8192 context and
nothing else, 2 or 4 GB, none to two cores. It gives the judge 2 GB,
--cache-ram 0 and cores by init's rule through modelsvc.SetProfiles, then
restarts its ready replicas. The replica count is left as it is. A judge
whose entry differs is left alone, and the log says so once.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 22:06:56 +02:00
Ilker ManapandClaude Opus 5.5 b8274d6278 A restart is not refused for room another restart is giving back
CI / test (push) Successful in 8m51s
On a 4-core single-machine install, `ctl model resources` was run on the
judge (to two cores) and on the starter (to none), then both restarted.
The judge's restart was refused: its replacement fit on no node even with
its old replica stopped, because the starter still held its two cores. A
restart now waits for other restarts under way (queued, retiring or
draining) when their going would make room, up to DrainGrace.

A drained replica was also kept until its model had enough live replicas,
and a replica being restarted that still answers did not count. So with
replicas restarted one after another, each drained one waited out the
five-minute grace, holding its room.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 21:33:20 +02:00
Ilker ManapandClaude Opus 5.5 8164260e3b Backlog: test VMs removed, Apertus on demand
CI / test (push) Failing after 6m12s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 21:18:46 +02:00
Ilker ManapandClaude Opus 5.5 fbb5df05b8 Backlog: what is open, in order; J, what the HA and single-machine tests found
CI / test (push) Successful in 8m59s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 21:07:57 +02:00
Ilker ManapandClaude Opus 5.5 af942aca27 v0.14.7: changelog and README
Release / build (push) Successful in 5m33s
Release / notes (push) Successful in 4s
CI / test (push) Successful in 10m30s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 20:11:26 +02:00
Ilker ManapandClaude Opus 5.5 3e7dd20bad One machine runs one judge, on up to four cores
CI / test (push) Successful in 9m1s
A second judge replica on the same machine shares its cores and is lost
with it, so it bought nothing but half the speed of each. On a 4-core,
8 GB single-machine install, one judge on four cores passed 20 of 20
requests at about 3.5 s each, judged in and out. Two judges on two cores
each had passed 19 of 20.

server init now serves one judge when this machine is the only node, and
two, on two cores each and spread over the nodes, where there will be
others (HA setups, or a control plane that is not a node).
--judge-replicas and the setup's Copies still choose. install.md gives
the measured times in place of "half a second".

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 19:35:42 +02:00
Ilker ManapandClaude Opus 5.5 f8206c3540 A judge reserves two cores and 2 GB, with no prompt cache
CI / test (push) Successful in 9m1s
Checked by a single-machine install of v0.14.6 on a 4-core, 8 GB VM. The
second judge fitted nowhere: v0.14.6 reserved 4 GB a judge, a figure from
one replica that grew to 3.2 GB through llama.cpp's prompt cache (up to
8 GB by default), which a judge never reuses. Once it fitted, every
judgment timed out. v0.14.6 gave a judge one core below 8 CPUs, and 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 a 5 s budget: 0 of 20 requests went through.

A judge now runs with --cache-ram 0, uses about 1.6 GB and reserves 2,
and reserves two cores (one only below four CPUs). On the same VM, 19 of
20 went through.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 19:22:33 +02:00
Ilker ManapandClaude Opus 5.5 0def6eaba1 CI: run on maas-packer, the runner that serves this Gitea, given 8 cores and 32 GB
CI / test (push) Successful in 9m1s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 18:15:58 +02:00
Ilker ManapandClaude Opus 5.5 645be2d278 CI: run again on the runner given 8 cores and 32 GB
CI / test (push) Canceled after 12s
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 17:49:44 +02:00
Ilker ManapandClaude Opus 5.5 25c4ec9f2a CI: a package that hangs says which test it hung in
CI / test (push) Failing after 30m4s
Every test run since c22528c ended at the job's 30-minute deadline with
nothing after internal/release, while the same tests pass with -race on
Linux, as a user and as root, in 5 minutes. A 12-minute test timeout makes
the hang name itself.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
2026-10-16 16:54:48 +02:00
57 changed files with 3205 additions and 78 deletions

No files matched your search

+5 -1
View File
@@ -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 ./...
+153 -3
View File
@@ -15,6 +15,41 @@ 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 (2026-10-16, after v0.14.7)
What is left, most needed first. The detail is in the item each names.
1. ~~**J1**~~ — done 2026-10-16: a restart no longer refused for room
another restart was giving back.
2. ~~**J5**~~ — done 2026-10-16: an upgrade refits an earlier install's
judge by itself.
3. ~~**I3**~~ — done 2026-10-16: `ctl bench overhead`, figures in the
README.
4. ~~**I4**~~ — done 2026-10-17: several keys for one outside provider.
5. ~~**Active Directory and Entra ID**~~ — done 2026-10-18: `auth.ldap`
sign-in and sync, Entra's 200-group overflow kept from taking roles.
Still to do: try it against a real domain controller (a Samba AD DC
on Proxmox): LDAPS, nested groups, a disabled user suspended by the
sync.
6. **J2** — a judge the leader adds by itself is sized as init's is. (S)
7. **J3** — HA setup puts the judges on nodes of their own. (S)
8. **J8** — a judge budget that fits the machine, measured at install. (S)
9. **F5** — large imports limited by the embedding engine's slots. (M)
10. **I1** — answers to similar questions from the cache. (M)
11. **J4** — the model mirror on more than one machine. (M)
12. **I2** — the customer's MCP servers behind llmgw. (L)
13. Smaller leftovers: audit before/after values, a tenant-admin role and
forms hidden from viewers (C1); Parquet and a scheduled push for usage
export (C4); English errors on Turkish pages (C24); H1 tried on the PoC;
licences in video and audio (H5); nginx in front of several control
planes (J7).
Not code, the operator's to decide or do: the PoC on v0.14.7 (it runs a
v0.14.5 dev build; no hurry); where Alertmanager sends (its receiver is
"none"). Settled 2026-10-16: Apertus 8B is on demand (min 0); the HA test
VMs 157–159, 165, 166 and the e2e VMs 151, 153–156 (a v0.11 HA test from
2026-10-03, idle since, holding 10.0.0.160 with VRRP id 51) were removed.
---
## A. Security — do first
@@ -1429,6 +1464,14 @@ console show and add them.
- 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)
@@ -1485,7 +1528,16 @@ in every client with its own credentials and nobody seeing what was called.
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)
### 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.
@@ -1500,7 +1552,13 @@ answered with a guess.
- 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)
### 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.
@@ -1513,9 +1571,101 @@ answering 429 makes the whole fallback answer 429.
- 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)
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.
### 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.
## 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.
+94 -1
View File
@@ -18,12 +18,105 @@ 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; v0.14.4: 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) |
| v0.14.0 | — | 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) |
| 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
APIs change between minors without a deprecation period.
## 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
+28 -1
View File
@@ -40,7 +40,7 @@ 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; v0.14.6: setup that explains every part and lets it be chosen, the gateway on all three HA machines) |
| 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) |
| 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
@@ -257,6 +257,33 @@ 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
+21
View File
@@ -9054,6 +9054,13 @@
"type": "string",
"description": "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": {
"type": "array",
"items": {
"$ref": "#/components/schemas/catalog.ExternalKey"
},
"description": "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": {
"type": "string",
"description": "Peer names the federated llmgw cluster this model is offered by, for a model that federation added (`llmgw ctl federation add`) rather than one written by hand. Such a model is named PEER/MODEL, comes and goes with the peer's own list, and is sent only while the peer is up."
@@ -9072,6 +9079,20 @@
}
}
},
"catalog.ExternalKey": {
"type": "object",
"description": "ExternalKey is one of a provider's keys.",
"properties": {
"file": {
"type": "string",
"description": "File holds the key, on the control plane, as APIKeyFile does."
},
"weight": {
"type": "number",
"description": "Weight is its share of the requests against the others'; zero is 1."
}
}
},
"catalog.GPUNeed": {
"type": "object",
"description": "GPUNeed is what card a profile needs.",
+111 -11
View File
@@ -152,9 +152,9 @@ 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.
@@ -187,15 +187,19 @@ 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`. 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.
@@ -1319,6 +1323,26 @@ 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`.
@@ -2517,6 +2541,82 @@ 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
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:
```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.
- **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`,
+150
View File
@@ -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) |
@@ -105,10 +106,13 @@ 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) |
@@ -144,6 +148,20 @@ 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) |
| `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) |
@@ -616,6 +634,15 @@ 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`
@@ -1103,6 +1130,15 @@ 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 +1171,20 @@ 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`
@@ -1401,6 +1451,106 @@ 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.
### `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`
+3
View File
@@ -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
@@ -45,6 +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
@@ -57,6 +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
+22
View File
@@ -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=
+63
View File
@@ -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)
}
}
+36 -1
View File
@@ -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)
+1
View File
@@ -150,6 +150,7 @@ func newBenchCmd(client func() *ctlClient, asJSON *bool) *cobra.Command {
f.DurationVar(&cfg.SLO.TPOTp99, "tpot-p99", cfg.SLO.TPOTp99, "per-output-token target")
f.BoolVar(&save, "save", false,
"record the capacity against the model and profile, so the number outlives this terminal")
cmd.AddCommand(newBenchOverheadCmd(asJSON))
return cmd
}
+421
View File
@@ -0,0 +1,421 @@
package cli
import (
"bytes"
"context"
"encoding/json"
"fmt"
"io"
"log/slog"
"net"
"net/http"
"os"
"os/exec"
"path/filepath"
"runtime"
"slices"
"sort"
"strings"
"sync"
"syscall"
"text/tabwriter"
"time"
"github.com/spf13/cobra"
"gitea.mynodes.xyz/ilker/llmgw/internal/config"
"gitea.mynodes.xyz/ilker/llmgw/internal/guard"
"gitea.mynodes.xyz/ilker/llmgw/internal/router"
storepkg "gitea.mynodes.xyz/ilker/llmgw/internal/store"
)
// What llmgw adds to a request, measured: the same load straight at an
// engine that answers at once, and through the gateway in front of it —
// bare, with authentication and usage accounting, and with the guardrails'
// pattern checks besides. The engine's own time is the same on both paths,
// so what differs is the gateway's.
//
// Everything runs in this process, on loopback: no cluster, no GPU, and the
// numbers are this machine's. On a machine shared with other work (a VM on
// an oversubscribed host) they move with it; compare modes run together,
// not figures from different runs.
// overheadMode is one way a request reaches the engine.
type overheadMode struct {
name string
gateway bool
auth, guard bool
}
var overheadModes = []overheadMode{
{name: "direct"},
{name: "gateway", gateway: true},
{name: "gateway+auth", gateway: true, auth: true},
{name: "gateway+auth+guard", gateway: true, auth: true, guard: true},
}
// overheadStep is one mode at one request rate.
type overheadStep struct {
Mode string `json:"mode"`
Rate int `json:"rate"`
Sent int `json:"sent"`
Failed int `json:"failed"`
P50 time.Duration `json:"p50"`
P99 time.Duration `json:"p99"`
AddedP50 time.Duration `json:"added_p50"`
AddedP99 time.Duration `json:"added_p99"`
// CPUPer1k is the process's CPU time per thousand requests beyond what
// the direct mode spent at the same rate: the gateway's.
CPUPer1k time.Duration `json:"cpu_per_1k"`
cpu time.Duration
}
func newBenchOverheadCmd(asJSON *bool) *cobra.Command {
var (
rates []int
each time.Duration
)
cmd := &cobra.Command{
Use: "overhead",
Short: "Measure what the gateway adds to a request, against an engine that answers at once",
Long: "Runs an engine that answers at once and a gateway in front of it, in this\n" +
"process on loopback, and sends the same load at each rate straight to the\n" +
"engine and through the gateway: bare, with authentication and usage\n" +
"accounting, and with the guardrails' pattern checks too. Reports each\n" +
"request's latency at p50 and p99, what the gateway added to it, and the\n" +
"gateway's CPU time per thousand requests.\n\n" +
"Needs no cluster. The figures are this machine's: on one shared with other\n" +
"work they move with it, so compare the modes of one run, and note the\n" +
"machine's load beside any figure you keep.",
Args: cobra.NoArgs,
RunE: func(cmd *cobra.Command, _ []string) error {
steps, err := runOverhead(cmd.Context(), rates, each)
if err != nil {
return err
}
if *asJSON {
return json.NewEncoder(cmd.OutOrStdout()).Encode(steps)
}
printOverhead(cmd.OutOrStdout(), steps)
return nil
},
}
f := cmd.Flags()
f.IntSliceVar(&rates, "rates", []int{100, 500, 1000, 2000}, "requests a second to send, one step each")
f.DurationVar(&each, "each", 3*time.Second, "how long each step sends for")
return cmd
}
func printOverhead(w io.Writer, steps []overheadStep) {
tw := tabwriter.NewWriter(w, 0, 0, 2, ' ', 0)
fmt.Fprintln(tw, "MODE\tRATE/S\tSENT\tFAILED\tP50\tP99\t+P50\t+P99\tCPU/1K REQ")
for _, s := range steps {
added := func(d time.Duration) string {
if s.Mode == "direct" {
return "-"
}
return msOf(d)
}
cpu := "-"
if s.Mode != "direct" {
cpu = msOf(s.CPUPer1k)
}
fmt.Fprintf(tw, "%s\t%d\t%d\t%d\t%s\t%s\t%s\t%s\t%s\n", s.Mode, s.Rate, s.Sent, s.Failed,
msOf(s.P50), msOf(s.P99), added(s.AddedP50), added(s.AddedP99), cpu)
}
_ = tw.Flush()
fmt.Fprintf(w, "\n%d CPUs, %s. Loopback, in one process: the figures are this machine's.\n",
runtime.NumCPU(), loadAverage())
}
func msOf(d time.Duration) string { return fmt.Sprintf("%.2fms", float64(d)/float64(time.Millisecond)) }
// loadAverage is the machine's, where it can be read, to keep beside the
// figures.
func loadAverage() string {
if b, err := os.ReadFile("/proc/loadavg"); err == nil {
if f := strings.Fields(string(b)); len(f) >= 3 {
return "load " + strings.Join(f[:3], " ")
}
}
// macOS: "{ 1.23 1.10 0.98 }".
if b, err := exec.Command("sysctl", "-n", "vm.loadavg").Output(); err == nil {
if f := strings.Fields(strings.Trim(strings.TrimSpace(string(b)), "{}")); len(f) >= 3 {
return "load " + strings.Join(f[:3], " ")
}
}
return "load not read"
}
// runOverhead measures every mode at every rate.
func runOverhead(ctx context.Context, rates []int, each time.Duration) ([]overheadStep, error) {
engine, err := instantEngine()
if err != nil {
return nil, err
}
defer engine.Close()
dir, err := os.MkdirTemp("", "llmgw-overhead-")
if err != nil {
return nil, err
}
defer os.RemoveAll(dir)
type target struct {
url, key string
stop func()
}
targets := map[string]target{}
defer func() {
for _, t := range targets {
if t.stop != nil {
t.stop()
}
}
}()
for _, m := range overheadModes {
if !m.gateway {
targets[m.name] = target{url: engine.URL}
continue
}
url, key, stop, err := overheadGateway(ctx, engine.URL, filepath.Join(dir, m.name+".db"), m)
if err != nil {
return nil, fmt.Errorf("%s: %w", m.name, err)
}
targets[m.name] = target{url: url, key: key, stop: stop}
}
client := &http.Client{Transport: &http.Transport{MaxIdleConns: 1024, MaxIdleConnsPerHost: 1024}, Timeout: 30 * time.Second}
var out []overheadStep
for _, rate := range rates {
var direct overheadStep
for _, m := range overheadModes {
t := targets[m.name]
s := loadStep(ctx, client, t.url, t.key, rate, each)
s.Mode = m.name
if m.name == "direct" {
direct = s
} else {
s.AddedP50, s.AddedP99 = max(0, s.P50-direct.P50), max(0, s.P99-direct.P99)
if s.Sent > 0 {
s.CPUPer1k = max(0, s.cpu-direct.cpu) * 1000 / time.Duration(s.Sent)
}
}
out = append(out, s)
}
}
return out, nil
}
// loadStep sends rate requests a second for d, each in its own goroutine,
// and measures each one's latency and the process's CPU meanwhile.
func loadStep(ctx context.Context, client *http.Client, url, key string, rate int, d time.Duration) overheadStep {
s := overheadStep{Rate: rate}
body := []byte(`{"model":"bench","max_tokens":8,"messages":[{"role":"user","content":"What is the capital of France?"}]}`)
var (
mu sync.Mutex
lat []time.Duration
wg sync.WaitGroup
total = int(float64(rate) * d.Seconds())
)
interval := time.Second / time.Duration(max(rate, 1))
cpu0 := processCPU()
start := time.Now()
for i := range total {
if ctx.Err() != nil {
break
}
if wait := time.Until(start.Add(time.Duration(i) * interval)); wait > 0 {
time.Sleep(wait)
}
wg.Add(1)
go func() {
defer wg.Done()
req, _ := http.NewRequestWithContext(ctx, http.MethodPost, url+"/v1/chat/completions", bytes.NewReader(body))
req.Header.Set("Content-Type", "application/json")
if key != "" {
req.Header.Set("Authorization", "Bearer "+key)
}
t := time.Now()
resp, err := client.Do(req)
ok := err == nil && resp.StatusCode == http.StatusOK
if resp != nil {
_, _ = io.Copy(io.Discard, resp.Body)
resp.Body.Close()
}
el := time.Since(t)
mu.Lock()
s.Sent++
if ok {
lat = append(lat, el)
} else {
s.Failed++
}
mu.Unlock()
}()
}
wg.Wait()
s.cpu = processCPU() - cpu0
s.P50, s.P99 = percentile(lat, 0.50), percentile(lat, 0.99)
return s
}
func percentile(ds []time.Duration, p float64) time.Duration {
if len(ds) == 0 {
return 0
}
ds = slices.Clone(ds)
sort.Slice(ds, func(i, j int) bool { return ds[i] < ds[j] })
i := int(p * float64(len(ds)-1))
return ds[i]
}
// processCPU is the CPU time this process has used, user and system.
func processCPU() time.Duration {
var ru syscall.Rusage
if err := syscall.Getrusage(syscall.RUSAGE_SELF, &ru); err != nil {
return 0
}
return time.Duration(ru.Utime.Nano() + ru.Stime.Nano())
}
// instantEngine is an OpenAI-compatible engine that answers at once, with
// the metrics the router scrapes.
func instantEngine() (*httpServer, error) {
mux := http.NewServeMux()
mux.HandleFunc("GET /metrics", func(w http.ResponseWriter, _ *http.Request) {
fmt.Fprint(w, "# TYPE vllm:num_requests_running gauge\nvllm:num_requests_running 0\n"+
"# TYPE vllm:num_requests_waiting gauge\nvllm:num_requests_waiting 0\n"+
"# TYPE vllm:kv_cache_usage_perc gauge\nvllm:kv_cache_usage_perc 0.1\n")
})
mux.HandleFunc("GET /health", func(w http.ResponseWriter, _ *http.Request) { w.WriteHeader(http.StatusOK) })
mux.HandleFunc("POST /v1/chat/completions", func(w http.ResponseWriter, r *http.Request) {
_, _ = io.Copy(io.Discard, r.Body)
w.Header().Set("Content-Type", "application/json")
fmt.Fprint(w, `{"id":"b","object":"chat.completion","model":"bench","choices":[{"index":0,"message":{"role":"assistant","content":"Paris."},"finish_reason":"stop"}],"usage":{"prompt_tokens":14,"completion_tokens":2,"total_tokens":16}}`)
})
return serveLoopback(mux)
}
// httpServer is a handler served on a loopback port.
type httpServer struct {
URL string
srv *http.Server
}
func (s *httpServer) Close() { _ = s.srv.Close() }
func serveLoopback(h http.Handler) (*httpServer, error) {
ln, err := net.Listen("tcp", "127.0.0.1:0")
if err != nil {
return nil, err
}
srv := &http.Server{Handler: h, ReadHeaderTimeout: 10 * time.Second}
go func() { _ = srv.Serve(ln) }()
return &httpServer{URL: "http://" + ln.Addr().String(), srv: srv}, nil
}
// overheadGateway starts a gateway in front of the engine as the mode asks,
// and returns its address and, with auth, a key.
func overheadGateway(ctx context.Context, engineURL, storePath string, m overheadMode) (string, string, func(), error) {
cfg := config.Default()
cfg.Listen = "127.0.0.1:0"
cfg.Router.ScrapeInterval = 200 * time.Millisecond
cfg.Endpoints = []config.Endpoint{{ID: "bench", URL: engineURL, Models: []string{"bench"}}}
key := ""
if m.auth {
var err error
if key, err = overheadKey(ctx, storePath); err != nil {
return "", "", nil, err
}
cfg.Auth.Enabled, cfg.Auth.Store = true, storePath
// Limits would measure themselves refusing: off.
cfg.Auth.DefaultRPM, cfg.Auth.DefaultTPM = 0, 0
}
rt, err := router.New(cfg, slog.New(slog.DiscardHandler))
if err != nil {
return "", "", nil, err
}
if m.guard {
g, err := baselineGuard()
if err != nil {
return "", "", nil, err
}
rt.SetGuard(g)
}
rctx, cancel := context.WithCancel(ctx)
rt.Start(rctx)
srv, err := serveLoopback(rt.Handler())
if err != nil {
cancel()
return "", "", nil, err
}
stop := func() { srv.Close(); cancel(); rt.Stop() }
// Ready once a request goes through: the endpoint is scraped first.
deadline := time.Now().Add(10 * time.Second)
for {
body := `{"model":"bench","messages":[{"role":"user","content":"hi"}]}`
req, _ := http.NewRequestWithContext(ctx, http.MethodPost, srv.URL+"/v1/chat/completions", strings.NewReader(body))
req.Header.Set("Content-Type", "application/json")
if key != "" {
req.Header.Set("Authorization", "Bearer "+key)
}
resp, err := http.DefaultClient.Do(req)
if err == nil {
_, _ = io.Copy(io.Discard, resp.Body)
resp.Body.Close()
if resp.StatusCode == http.StatusOK {
return srv.URL, key, stop, nil
}
}
if time.Now().After(deadline) {
stop()
status := "no answer"
if err == nil {
status = resp.Status
}
return "", "", nil, fmt.Errorf("the gateway did not serve within 10s (%s)", status)
}
time.Sleep(100 * time.Millisecond)
}
}
// overheadKey makes a store with one project and a key in it.
func overheadKey(ctx context.Context, path string) (string, error) {
st, err := storepkg.Open(ctx, path)
if err != nil {
return "", err
}
defer st.Close()
org, err := st.CreateOrg(ctx, "bench")
if err != nil {
return "", err
}
team, err := st.CreateTeam(ctx, org, "bench")
if err != nil {
return "", err
}
project, err := st.CreateProject(ctx, team, "bench")
if err != nil {
return "", err
}
secret, k, err := storepkg.NewKey("key-bench", "bench", project)
if err != nil {
return "", err
}
k.CreatedAt = time.Now()
return secret, st.CreateKey(ctx, k)
}
// baselineGuard is the policy init installs, without its judge: the pattern
// checks every request passes.
func baselineGuard() (guard.Guard, error) {
m, err := config.Parse([]byte(guardSection(false)))
if err != nil {
return nil, err
}
cfg, err := config.FromMap(m)
if err != nil {
return nil, err
}
return guard.NewGate(cfg.Guard, nil, slog.New(slog.DiscardHandler))
}
+33
View File
@@ -0,0 +1,33 @@
package cli
import (
"bytes"
"strings"
"testing"
"time"
)
// The short version, so CI sees the gateway's overhead on every change: every
// mode answers every request, and what the gateway adds is measured, not
// guessed. No bound is asserted on it, as CI's machines vary; the table is
// in the test's log.
func TestTheGatewaysOverheadIsMeasured(t *testing.T) {
steps, err := runOverhead(t.Context(), []int{50}, time.Second)
if err != nil {
t.Fatal(err)
}
if len(steps) != len(overheadModes) {
t.Fatalf("%d steps for %d modes", len(steps), len(overheadModes))
}
for _, s := range steps {
if s.Sent == 0 || s.Failed > 0 || s.P50 <= 0 || s.P99 < s.P50 {
t.Errorf("%s: sent %d failed %d p50 %s p99 %s", s.Mode, s.Sent, s.Failed, s.P50, s.P99)
}
}
var b bytes.Buffer
printOverhead(&b, steps)
if !strings.Contains(b.String(), "gateway+auth+guard") {
t.Errorf("table:\n%s", b.String())
}
t.Logf("\n%s", b.String())
}
+99
View File
@@ -32,6 +32,19 @@ type judgeKeeper struct {
// tried is when each judge was last asked for, so one that cannot be
// fetched is not asked for again on every look.
tried map[string]time.Time
// What refitting an earlier install's judge needs: the catalog to change
// its profile in, the replicas to restart, and the nodes' CPUs. Any nil
// leaves judges as they are.
profiles interface {
SetProfiles(ctx context.Context, name string, profiles []catalog.Profile, by string) (catalog.Model, error)
}
restart func(ids ...string) error
replicas func(model string) []string
nodeCPUs func() []int
// looked is each judge already refitted or found changed by hand, so
// each is said once.
looked map[string]bool
}
// judgeRetry is how long a judge whose serving failed waits before it is
@@ -69,6 +82,7 @@ func (k *judgeKeeper) check() {
have := map[string]bool{}
for _, m := range k.models() {
have[m.Name] = true
k.refit(m)
}
busy := map[string]bool{}
for _, j := range k.serve.Jobs() {
@@ -114,3 +128,88 @@ func (k *judgeKeeper) check() {
"judge", name, "repo", kj.Repo, "file", kj.File, "job", job.ID)
}
}
// refit moves a judge an earlier `server init` installed to what init
// installs now, and restarts its replicas onto it. Those judges reserved no
// cores or one, 2 or 4 GB, and ran llama.cpp's prompt cache: on fewer than
// eight cores every judgment missed its 5 s budget and fail: closed refused
// every request, until fixed by hand. A judge whose profile differs in any
// other way was set by somebody, and is left as it is, said once.
func (k *judgeKeeper) refit(m catalog.Model) {
if k.profiles == nil || k.restart == nil || k.replicas == nil || m.Name != defaultJudge || !m.IsGuard() {
return
}
if k.looked == nil {
k.looked = map[string]bool{}
}
if k.looked[m.Name] {
return
}
k.looked[m.Name] = true
if !earlierInitJudge(m) {
if len(m.Profiles) != 1 || !slices.Contains(m.Profiles[0].Args, "--cache-ram") {
k.log.Info("the judge's profile is not one init wrote; left as it is",
"judge", m.Name, "see", "llmgw ctl model resources "+m.Name)
}
return
}
var cpus []int
if k.nodeCPUs != nil {
cpus = k.nodeCPUs()
}
p := m.Profiles[0]
p.MinMemoryGB = judgeMemoryGB
p.CPUCores = refitCores(m.Scaling.Min, cpus)
p.Args = []string{"--ctx-size", "8192", "--cache-ram", "0"}
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
defer cancel()
if _, err := k.profiles.SetProfiles(ctx, m.Name, []catalog.Profile{p}, "llmgw (the judge an earlier init installed)"); err != nil {
k.log.Warn("could not refit the judge an earlier init installed", "judge", m.Name, "err", err)
k.looked[m.Name] = false
return
}
ids := k.replicas(m.Name)
if len(ids) > 0 {
if err := k.restart(ids...); err != nil {
k.log.Warn("refitted the judge, but could not restart it", "judge", m.Name, "err", err)
}
}
k.log.Info("refitted the judge an earlier init installed", "judge", m.Name,
"memory_gb", p.MinMemoryGB, "cpu_cores", p.CPUCores, "replicas_restarted", len(ids))
}
// earlierInitJudge is whether a judge's entry is exactly what `server init`
// before v0.14.7 wrote: one llama.cpp profile on CPUs, an 8192 context and
// no other args, 2 or 4 GB, none to two cores.
func earlierInitJudge(m catalog.Model) bool {
if len(m.Profiles) != 1 {
return false
}
p := m.Profiles[0]
return p.Name == "cpu" && p.Engine == "llamacpp" && p.GPUClass == "" && p.GPU == nil &&
len(p.NodeSelector) == 0 && slices.Equal(p.Args, []string{"--ctx-size", "8192"}) &&
(p.MinMemoryGB == 2 || p.MinMemoryGB == 4) && p.CPUCores >= 0 && p.CPUCores <= 2
}
// refitCores is a judge replica's cores by init's rule: up to four for the
// one judge of a cluster, two each where there are more, fewer where the
// smallest node cannot hold them all.
func refitCores(replicas int, cpus []int) int {
smallest := 0
for _, c := range cpus {
if c > 0 && (smallest == 0 || c < smallest) {
smallest = c
}
}
if replicas <= 1 {
if smallest == 0 {
return judgeCores
}
return min(smallest, judgeCoresAlone)
}
cores := judgeCores
if len(cpus) == 1 && smallest > 0 && replicas*cores > smallest {
cores = max(1, smallest/replicas)
}
return cores
}
+79
View File
@@ -1,6 +1,7 @@
package cli
import (
"context"
"errors"
"log/slog"
"slices"
@@ -124,3 +125,81 @@ func TestCategoriesFollowTheJudges(t *testing.T) {
t.Errorf("no categories (every one counts) became %v", got)
}
}
type fakeProfiles struct {
set map[string][]catalog.Profile
}
func (f *fakeProfiles) SetProfiles(_ context.Context, name string, ps []catalog.Profile, _ string) (catalog.Model, error) {
if f.set == nil {
f.set = map[string][]catalog.Profile{}
}
f.set[name] = ps
return catalog.Model{Name: name, Profiles: ps}, nil
}
// A judge that v0.14.6 or earlier's init installed is moved to today's
// figures, once, and its replicas restarted onto them; one somebody changed
// is left alone.
func TestAnEarlierInstallsJudgeIsRefittedOnce(t *testing.T) {
lib := guard.Config{Policies: map[string]guard.Policy{
"baseline": {Judge: &guard.JudgePolicy{Model: "qwen3guard-0.6b", Input: "block"}},
}}
earlier := guardModel("http://10.0.0.5:8081/qwen3guard-0.6b/guard/g.gguf", 2, 1)
earlier.Profiles[0].MinMemoryGB, earlier.Profiles[0].Args = 4, []string{"--ctx-size", "8192"}
inCatalog := []catalog.Model{earlier}
fp := &fakeProfiles{}
var restarted []string
k := &judgeKeeper{
library: func() guard.Config { return lib },
models: func() []catalog.Model { return inCatalog },
serve: &fakeServe{},
log: slog.New(slog.DiscardHandler),
profiles: fp,
restart: func(ids ...string) error { restarted = append(restarted, ids...); return nil },
replicas: func(string) []string { return []string{"j-1", "j-2"} },
nodeCPUs: func() []int { return []int{4} },
}
k.check()
got := fp.set["qwen3guard-0.6b"]
if len(got) != 1 || got[0].MinMemoryGB != 2 || got[0].CPUCores != 2 || !slices.Contains(got[0].Args, "--cache-ram") {
t.Fatalf("refitted to %+v", got)
}
if !slices.Equal(restarted, []string{"j-1", "j-2"}) {
t.Errorf("restarted %v", restarted)
}
k.check()
if len(restarted) != 2 {
t.Error("refitted twice")
}
// Changed by somebody: an extra arg. Left as it is.
mine := guardModel("http://x/g.gguf", 2, 3)
mine.Profiles[0].Args = []string{"--ctx-size", "16384"}
inCatalog = []catalog.Model{mine}
fp.set, k.looked = nil, nil
k.check()
if fp.set != nil {
t.Errorf("a judge set by hand was changed: %+v", fp.set)
}
}
func TestAJudgeIsRefittedToTheCoresInitWouldGive(t *testing.T) {
for _, c := range []struct {
replicas int
cpus []int
want int
}{
{1, []int{4}, 4},
{1, []int{2}, 2},
{1, []int{16}, 4},
{2, []int{4}, 2},
{2, []int{2}, 1},
{2, []int{4, 32}, 2},
{2, nil, 2},
} {
if got := refitCores(c.replicas, c.cpus); got != c.want {
t.Errorf("%d judges on %v: %d cores, want %d", c.replicas, c.cpus, got, c.want)
}
}
}
+19
View File
@@ -593,6 +593,25 @@ func newServerCmd() *cobra.Command {
},
gpus: func() []adminapi.ClassUse { return adminAPI.Summary().GPUs },
log: log,
// An earlier install's judge, moved to today's figures.
profiles: catalogSvc,
restart: admin.Reconciler().Restart,
replicas: func(model string) []string {
var ids []string
for _, r := range admin.Reconciler().ReadyReplicas() {
if r.Model == model {
ids = append(ids, r.ID)
}
}
return ids
},
nodeCPUs: func() []int {
var out []int
for _, n := range admin.Nodes() {
out = append(out, n.CPUCount)
}
return out
},
}
go keeper.run(ctx, 30*time.Second)
}
+15 -1
View File
@@ -194,6 +194,9 @@ func (in *serverInit) judgeFor() judgeSpec {
}
if j.Replicas <= 0 {
j.Replicas = 2
if in.singleMachine() {
j.Replicas = 1
}
}
if j.Budget <= 0 {
j.Budget = 5 * time.Second
@@ -201,6 +204,14 @@ func (in *serverInit) judgeFor() judgeSpec {
return j
}
// singleMachine is an install where this machine is the only node there
// will be, as far as init knows: a second judge replica there shares the
// same cores and is lost with the same machine, so it buys nothing but
// half the speed of each.
func (in *serverInit) singleMachine() bool {
return !in.noNode && !in.standbys && len(in.patroni) == 0 && len(in.patroniNodes) == 0
}
// The configuration's limits, as init writes them unless told otherwise.
const (
defaultRPM = 60
@@ -346,7 +357,10 @@ func (in *serverInit) initFlags() []string {
if in.noGuardModel {
add("--no-guard-model")
} else {
j, d := in.judgeFor(), (&serverInit{}).judgeFor()
// Against the defaults for an install of the same shape: one judge
// on one machine, two otherwise.
shape := serverInit{noNode: in.noNode, standbys: in.standbys, patroni: in.patroni, patroniNodes: in.patroniNodes}
j, d := in.judgeFor(), shape.judgeFor()
if j.Fail != d.Fail {
add("--judge-fail", j.Fail)
}
+40 -7
View File
@@ -183,12 +183,17 @@ func TestTheJudgesAndTheStarterShareTheCPUsRatherThanEachTakingAll(t *testing.T)
in serverInit
judge, starter int
}{
{serverInit{cpus: 4}, 1, 2},
{serverInit{cpus: 8}, 2, 4},
{serverInit{cpus: 32}, 2, 8},
{serverInit{cpus: 2}, 1, 0}, // nothing left: the starter shares rather than fits nowhere
// One machine: one judge, on up to four cores; the starter has the rest.
{serverInit{cpus: 4}, 4, 0},
{serverInit{cpus: 8}, 4, 4},
{serverInit{cpus: 32}, 4, 8},
{serverInit{cpus: 2}, 2, 0},
{serverInit{cpus: 1}, 0, 0},
{serverInit{cpus: 4, noGuardModel: true}, 1, 4},
{serverInit{cpus: 4, noGuardModel: true}, judgeCores, 4}, // no judge: all for the starter
// Two asked for on one machine: two cores each, one each where two
// would not fit both.
{serverInit{cpus: 8, judge: judgeSpec{Replicas: 2}}, 2, 4},
{serverInit{cpus: 2, judge: judgeSpec{Replicas: 2}}, 1, 0},
{serverInit{cpus: 8, judge: judgeSpec{Replicas: 3}}, 2, 2},
{serverInit{cpus: 4, noNode: true}, 2, 0}, // on nodes not known yet
} {
@@ -198,11 +203,39 @@ func TestTheJudgesAndTheStarterShareTheCPUsRatherThanEachTakingAll(t *testing.T)
}
}
m := guardModel("http://x/g.gguf", 2, 1)
if p := m.Profiles[0]; p.CPUCores != 1 || p.MinMemoryGB != 4 {
t.Errorf("judge profile: %d cores, %d GB", p.CPUCores, p.MinMemoryGB)
if p := m.Profiles[0]; p.CPUCores != 1 || p.MinMemoryGB != 2 || !slices.Contains(p.Args, "--cache-ram") {
t.Errorf("judge profile: %d cores, %d GB, args %v", p.CPUCores, p.MinMemoryGB, p.Args)
}
s := starterModel(starterChoices[0], defaultContext, 2, "http://x/q.gguf", []string{"6000-24g"}, true)
if cpu := s.Profiles[len(s.Profiles)-1]; cpu.Name != "cpu" || cpu.CPUCores != 2 || s.Profiles[0].CPUCores != 0 {
t.Errorf("starter: GPU profile %d cores, CPU %+v", s.Profiles[0].CPUCores, cpu)
}
}
// A second judge on the same machine shares its cores and is lost with it:
// one machine has one, more machines two.
func TestOneMachineHasOneJudgeAndMoreHaveTwo(t *testing.T) {
for _, c := range []struct {
in serverInit
want int
}{
{serverInit{}, 1},
{serverInit{noNode: true}, 2},
{serverInit{patroni: []string{"a@db1", "b@db2"}}, 2},
{serverInit{standbys: true}, 2},
{serverInit{judge: judgeSpec{Replicas: 3}}, 3},
} {
if got := c.in.judgeFor().Replicas; got != c.want {
t.Errorf("%+v: %d judges, want %d", c.in, got, c.want)
}
}
// The defaults are left out of the command; a choice against them is in.
ha := serverInit{noNode: true, patroni: []string{"a@db1", "b@db2"}, judge: judgeSpec{Replicas: 2}}
if f := strings.Join(ha.initFlags(), " "); strings.Contains(f, "--judge-replicas") {
t.Errorf("two judges on three machines named: %s", f)
}
one := serverInit{judge: judgeSpec{Replicas: 2}}
if f := strings.Join(one.initFlags(), " "); !strings.Contains(f, "--judge-replicas 2") {
t.Errorf("two judges on one machine not named: %s", f)
}
}
+32 -17
View File
@@ -496,10 +496,13 @@ func guardModel(source string, replicas, cores int) catalog.Model {
// judge that is not there refuses everything.
Scaling: catalog.Scaling{Min: replicas, Max: replicas},
Profiles: []catalog.Profile{{
// It uses about 3.2 GB with its context: reserving 2 filled nodes
// past their memory.
Name: "cpu", Engine: "llamacpp", Priority: 100, MinMemoryGB: 4, CPUCores: cores,
Args: []string{"--ctx-size", "8192"},
// About 1.6 GB with its context, which stays so with no prompt
// cache: llama.cpp keeps up to 8 GB of prompts by default, and a
// judge's, each a different text, are not asked again. With the
// cache one grew to 3.2 GB over a day of traffic; reserving 4 for
// that left a 7 GB machine room for only one of the two.
Name: "cpu", Engine: "llamacpp", Priority: 100, MinMemoryGB: judgeMemoryGB, CPUCores: cores,
Args: []string{"--ctx-size", "8192", "--cache-ram", "0"},
}},
}
}
@@ -716,14 +719,16 @@ type cpuSplit struct{ judge, starter int }
// the starter on one 4-core control plane all ran on all four: under load
// most judgments missed their budget and the requests were refused.
//
// When this machine is the node, by its own CPUs: a core a judge under 8,
// two from 8, and the starter what is left, up to 8, or no reservation at
// all when nothing is, rather than a starter that fits nowhere. Otherwise
// two a judge, on whichever node takes it, and the starter, on a node not
// known yet, none.
// A judge needs cores: a judgment of about 280 tokens took 8.0 s on one,
// 4.2 s on two and 2.3 s on four, against a 5 s budget. On one machine its
// one replica gets up to four; where there are two, two each (one each only
// where two would not fit both). The starter has what is left, up to 8, or
// no reservation at all when nothing is, rather than a starter that fits
// nowhere. A machine that is not the node reserves two a judge on whichever
// node takes it, and the starter, on a node not known yet, none.
func (in *serverInit) cpuPlan() cpuSplit {
if in.noNode {
return cpuSplit{judge: 2}
return cpuSplit{judge: judgeCores}
}
n := in.cpus
if n <= 0 {
@@ -736,13 +741,23 @@ func (in *serverInit) cpuPlan() cpuSplit {
if !in.noGuardModel {
judges = in.judgeFor().Replicas
}
judge := 1
if n >= 8 {
judge = 2
}
left := n - judges*judge
if left < 1 {
left = 0
judge := judgeCores
switch {
case judges == 1:
judge = min(n, judgeCoresAlone)
case judges > 0 && judges*judge > n:
judge = max(1, n/judges)
}
left := max(n-judges*judge, 0)
return cpuSplit{judge: judge, starter: min(left, 8)}
}
// judgeCores and judgeMemoryGB are what a judge replica needs: two cores
// for a judgment within its budget, and 2 GB for its 1.6 with no prompt
// cache. judgeCoresAlone is what the one judge of a single machine takes,
// past which a model this small hardly gets faster.
const (
judgeCores = 2
judgeCoresAlone = 4
judgeMemoryGB = 2
)
+1 -1
View File
@@ -282,7 +282,7 @@ func newInstallForm(in *serverInit, host string, found []string, ha bool) *insta
"The request is served unjudged and marked so in the logs. Nothing is refused for the judge's sake.",
}
f.judgeCopies = textSetupField("Copies", strconv.Itoa(j.Replicas), "2",
"Each takes about 1 GB and two CPU threads on whichever node has room.")
"On one machine one copy, on up to four cores; with more machines two, two cores each, on different nodes. Each reserves 2 GB.")
f.judgeBudget = textSetupField("Waits at most", durationYAML(j.Budget), "5s",
"A short text is judged in about two seconds on CPUs.")
judgeFields := []*setupField{f.judgeFail, f.judgeCopies, f.judgeBudget, headingField("Refuses")}
+2
View File
@@ -188,6 +188,8 @@ func newSetupModel(host string, ips []string) setupModel {
// nodes are added once it is up. Either can still be chosen.
in.noNode, in.noModel = true, true
in.standbys = true
// Two judges, on two of the three machines' nodes.
in.judge.Replicas = 2
return setupModel{
ref: &progRef{}, host: host, ips: ips, in: &in, choices: newInstallForm(&in, host, ips, true),
db: [2]dbMachine{newDBMachine(1), newDBMachine(2)}, spin: newSpinner(), width: 100, height: 30,
+9
View File
@@ -442,6 +442,9 @@ type Auth struct {
// OIDC is sign-in through an identity provider.
OIDC OIDC `koanf:"oidc"`
// LDAP is sign-in against a directory, Active Directory above all, and
// the accounts kept in step with it.
LDAP LDAP `koanf:"ldap"`
// SCIM is account provisioning by the identity provider.
SCIM SCIM `koanf:"scim"`
@@ -968,6 +971,12 @@ func (s Server) Validate() error {
if err := s.Auth.OIDC.Validate(); err != nil {
return err
}
if err := s.Auth.LDAP.Validate(); err != nil {
return err
}
if s.Auth.LDAP.Enabled() && !s.Auth.Enabled {
return fmt.Errorf("auth.ldap needs auth.enabled: it signs people in to the accounts the store keeps")
}
if err := s.AgentTLS.Validate(); err != nil {
return err
}
+137
View File
@@ -0,0 +1,137 @@
package config
import (
"errors"
"fmt"
"net/url"
"strings"
"time"
)
// LDAP is sign-in against a directory — Active Directory above all — with
// the directory's own passwords, and the accounts kept in step with it.
//
// auth:
// ldap:
// urls: [ldaps://dc1.corp.example.com, ldaps://dc2.corp.example.com]
// bind_dn: CN=llmgw,OU=Service Accounts,DC=corp,DC=example,DC=com
// bind_password_file: /etc/llmgw/secrets/ldap-bind
// base_dn: OU=Staff,DC=corp,DC=example,DC=com
// create_accounts: true
// roles: {admin: [llmgw-admins]}
// memberships: {ml-team: "team:Acme/ml"}
//
// A person signs in on the portal's form with their sAMAccountName, their
// userPrincipalName or their mail, and their directory password. The
// gateway finds them with the bind account, checks the password by binding
// as them, and reads their groups, nested ones included. Every sync_interval
// the leader reads the directory again: an account disabled or gone there
// is suspended here, its sessions ended and its API keys refused.
type LDAP struct {
// URLs are the directory's servers, tried in turn: ldaps:// (636), or
// ldap:// (389) with StartTLS.
URLs []string `koanf:"urls"`
// StartTLS upgrades an ldap:// connection before anything is sent. A
// password never crosses a connection without TLS: ldap:// without it
// is refused.
StartTLS bool `koanf:"start_tls"`
// CAFile is the authority the servers' certificates are checked against,
// when it is not the system's: an AD Certificate Services root.
CAFile string `koanf:"ca_file"`
// BindDN and the file holding its password: a service account that may
// read users and groups. Nothing is written to the directory.
BindDN string `koanf:"bind_dn"`
BindPasswordFile string `koanf:"bind_password_file"`
// BaseDN is where people are searched for, subtree included.
BaseDN string `koanf:"base_dn"`
// GroupBaseDN is where their groups are; empty is BaseDN's domain root.
GroupBaseDN string `koanf:"group_base_dn"`
// 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.
UserFilter string `koanf:"user_filter"`
// Label is what the form says to sign in with: "Sign in with your
// <label> account". Empty is "Active Directory".
Label string `koanf:"label"`
// 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.
CreateAccounts bool `koanf:"create_accounts"`
// 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.
Roles map[string][]string `koanf:"roles"`
// 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.
Memberships map[string]string `koanf:"memberships"`
// SyncInterval is how often the leader reads the directory. Zero is
// 15m; negative turns the sync off, leaving only sign-in.
SyncInterval time.Duration `koanf:"sync_interval"`
// Timeout bounds each round trip to a server. Zero is 10s.
Timeout time.Duration `koanf:"timeout"`
}
// Enabled reports whether directory sign-in is configured.
func (l LDAP) Enabled() bool { return len(l.URLs) > 0 }
// Validate says what in the configuration cannot work.
func (l LDAP) Validate() error {
if !l.Enabled() {
return nil
}
for _, raw := range l.URLs {
u, err := url.Parse(raw)
if err != nil || u.Host == "" {
return fmt.Errorf("auth.ldap.urls: %q is not an ldap:// or ldaps:// address", raw)
}
switch u.Scheme {
case "ldaps":
case "ldap":
if !l.StartTLS {
return fmt.Errorf("auth.ldap.urls: %q is plaintext: use ldaps://, or set start_tls; passwords would cross it", raw)
}
default:
return fmt.Errorf("auth.ldap.urls: %q is not an ldap:// or ldaps:// address", raw)
}
}
switch {
case strings.TrimSpace(l.BindDN) == "" || l.BindPasswordFile == "":
return errors.New("auth.ldap: bind_dn and bind_password_file are needed, for the account that finds people")
case strings.TrimSpace(l.BaseDN) == "":
return errors.New("auth.ldap: base_dn is needed, where people are searched for")
case l.UserFilter != "" && !(strings.HasPrefix(l.UserFilter, "(") && strings.HasSuffix(l.UserFilter, ")")):
return fmt.Errorf("auth.ldap.user_filter: %q is not an LDAP filter in parentheses", l.UserFilter)
}
for role := range l.Roles {
if !validStaffRole(role) {
return fmt.Errorf("auth.ldap.roles: %q is not viewer, operator, billing or admin", role)
}
}
for g, place := range l.Memberships {
if !validPlace(place) {
return fmt.Errorf("auth.ldap.memberships[%s]: %q is not org:NAME, team:ORG/TEAM or project:ORG/TEAM/PROJECT", g, place)
}
}
return nil
}
// Interval is SyncInterval with its default; zero means off.
func (l LDAP) Interval() time.Duration {
switch {
case l.SyncInterval < 0:
return 0
case l.SyncInterval == 0:
return 15 * time.Minute
}
return l.SyncInterval
}
// RoundTrip is Timeout with its default.
func (l LDAP) RoundTrip() time.Duration {
if l.Timeout <= 0 {
return 10 * time.Second
}
return l.Timeout
}
+18 -5
View File
@@ -72,21 +72,34 @@ func (o OIDC) Validate() error {
return errors.New("auth.oidc.client_id is required")
}
for role := range o.Roles {
switch role {
case "viewer", "operator", "billing", "admin":
default:
if !validStaffRole(role) {
return fmt.Errorf("auth.oidc.roles: %q is not viewer, operator, billing or admin", role)
}
}
for g, where := range o.Memberships {
kind, path, ok := strings.Cut(where, ":")
if !ok || path == "" || (kind != "org" && kind != "team" && kind != "project") {
if !validPlace(where) {
return fmt.Errorf("auth.oidc.memberships.%s is %q: org:NAME, team:ORG/TEAM or project:ORG/TEAM/PROJECT", g, where)
}
}
return nil
}
// validStaffRole is a role groups may give: viewer, operator, billing, admin.
func validStaffRole(role string) bool {
switch role {
case "viewer", "operator", "billing", "admin":
return true
}
return false
}
// validPlace is a place in the tree a group may give: org:NAME,
// team:ORG/TEAM or project:ORG/TEAM/PROJECT.
func validPlace(where string) bool {
kind, path, ok := strings.Cut(where, ":")
return ok && path != "" && (kind == "org" || kind == "team" || kind == "project")
}
// SCIM lets an identity provider provision accounts (/scim/v2): made,
// suspended and reinstated as they are there, with groups mapped by
// auth.oidc's roles and memberships.
+27
View File
@@ -1533,6 +1533,11 @@ type External struct {
// 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.
APIKeyFile string `json:"api_key_file,omitempty" koanf:"api_key_file"`
// 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.
APIKeyFiles []ExternalKey `json:"api_key_files,omitempty" koanf:"api_key_files"`
// Peer names the federated llmgw cluster this model is offered by, for
// a model that federation added (`llmgw ctl federation add`) rather
// than one written by hand. Such a model is named PEER/MODEL, comes and
@@ -1551,6 +1556,14 @@ type External struct {
Output string `json:"output,omitempty" koanf:"output"`
}
// ExternalKey is one of a provider's keys.
type ExternalKey struct {
// File holds the key, on the control plane, as APIKeyFile does.
File string `json:"file" koanf:"file"`
// Weight is its share of the requests against the others'; zero is 1.
Weight float64 `json:"weight,omitempty" koanf:"weight"`
}
// APIComfyUI is External.API for a ComfyUI server.
const APIComfyUI = "comfyui"
@@ -1568,6 +1581,20 @@ func (e *External) validate(model string, profiles int) error {
if strings.TrimSpace(e.Model) == "" {
return fmt.Errorf("model %q: external model needs the name the provider knows it by", model)
}
if e.APIKeyFile != "" && len(e.APIKeyFiles) > 0 {
return fmt.Errorf("model %q: api_key_file or api_key_files, not both", model)
}
for i, k := range e.APIKeyFiles {
if strings.TrimSpace(k.File) == "" {
return fmt.Errorf("model %q: api_key_files[%d] names no file", model, i)
}
if k.Weight < 0 {
return fmt.Errorf("model %q: api_key_files[%d] has a negative weight", model, i)
}
}
if len(e.APIKeyFiles) > 0 && e.Peer != "" {
return fmt.Errorf("model %q: a federated model is reached with its peer's key, not api_key_files", model)
}
switch e.API {
case "", "openai":
if e.Workflow != "" || e.Output != "" {
+20
View File
@@ -429,3 +429,23 @@ func TestAGuardModelSaysHowItIsReadAndAlwaysRuns(t *testing.T) {
t.Error("a judge is not critical by default")
}
}
func TestAProvidersKeysAreOneListOrOneFile(t *testing.T) {
ext := func(e External) Model {
e.URL, e.Model = "https://api.example.com", "m"
return Model{Name: "cloud", Source: "external", External: &e}
}
for _, bad := range []External{
{APIKeyFile: "/k", APIKeyFiles: []ExternalKey{{File: "/k2"}}},
{APIKeyFiles: []ExternalKey{{File: " "}}},
{APIKeyFiles: []ExternalKey{{File: "/k", Weight: -1}}},
{Peer: "p", APIKeyFiles: []ExternalKey{{File: "/k"}}},
} {
if _, err := New([]Model{ext(bad)}); err == nil {
t.Errorf("taken: %+v", bad)
}
}
if _, err := New([]Model{ext(External{APIKeyFiles: []ExternalKey{{File: "/a", Weight: 2}, {File: "/b"}}})}); err != nil {
t.Error(err)
}
}
+28 -12
View File
@@ -438,15 +438,6 @@ func (c *Controller) externalEndpoints() []endpoint.Info {
continue
}
}
key := ""
if f := m.External.APIKeyFile; f != "" {
b, err := secrets.Read(context.Background(), f)
if err != nil {
c.log.Warn("an external model's key could not be read; not offered", "model", m.Name, "err", err)
continue
}
key = strings.TrimSpace(string(b))
}
// A peer's endpoint says so in its id, which is what usage records
// as the replica that answered.
id := "external:" + m.Name
@@ -462,14 +453,39 @@ func (c *Controller) externalEndpoints() []endpoint.Info {
}
engine, workflow = endpoint.EngineComfyUI, w
}
out = append(out, endpoint.Info{
base := endpoint.Info{
ID: id, URL: strings.TrimRight(m.External.URL, "/"),
Models: append([]string{m.Name}, m.Aliases...), APIKey: key,
Models: append([]string{m.Name}, m.Aliases...),
Engine: engine, Pool: "external", Weight: 1,
UpstreamModel: m.External.Model,
Peer: m.External.Peer, PeerCluster: cluster, PeerTags: tags,
Workflow: workflow,
})
}
keys := m.External.APIKeyFiles
if len(keys) == 0 {
keys = []catalog.ExternalKey{{File: m.External.APIKeyFile}}
}
for i, k := range keys {
e := base
// One endpoint a key, numbered, so usage says which answered
// and a breaker stands for one key — never the key itself.
if len(m.External.APIKeyFiles) > 0 {
e.ID = fmt.Sprintf("%s#%d", id, i+1)
if k.Weight > 0 {
e.Weight = k.Weight
}
}
if k.File != "" {
b, err := secrets.Read(context.Background(), k.File)
if err != nil {
c.log.Warn("an external model's key could not be read; not offered", "model", m.Name,
"endpoint", e.ID, "err", err)
continue
}
e.APIKey = strings.TrimSpace(string(b))
}
out = append(out, e)
}
}
return out
}
+25
View File
@@ -73,3 +73,28 @@ func TestAnExternalModelIsPublishedWithItsKeyAndNothingIsPlacedForIt(t *testing.
}
}
}
// Several keys for one provider: an endpoint each, numbered, with its
// weight; one whose file cannot be read is left out and the others stay.
func TestEachOfAProvidersKeysIsAnEndpointOfItsOwn(t *testing.T) {
dir := t.TempDir()
a, c := filepath.Join(dir, "a.key"), filepath.Join(dir, "c.key")
_ = os.WriteFile(a, []byte("sk-a\n"), 0o600)
_ = os.WriteFile(c, []byte("sk-c"), 0o600)
cat, err := catalog.New([]catalog.Model{{Name: "cloud", Source: "external",
External: &catalog.External{URL: "https://api.example.com", Model: "gpt-4o-mini",
APIKeyFiles: []catalog.ExternalKey{{File: a, Weight: 3}, {File: filepath.Join(dir, "gone")}, {File: c}}}}})
if err != nil {
t.Fatal(err)
}
sink := &recordSink{}
New(DefaultConfig(), catalog.NewLive(cat), noDispatch{}, sink, slog.New(slog.DiscardHandler)).Publish(t.Context())
got := map[string]endpoint.Info{}
for _, e := range sink.infos {
got[e.ID] = e
}
if len(got) != 2 || got["external:cloud#1"].APIKey != "sk-a" || got["external:cloud#1"].Weight != 3 ||
got["external:cloud#3"].APIKey != "sk-c" || got["external:cloud#3"].Weight != 1 {
t.Errorf("published %+v", sink.infos)
}
}
@@ -504,7 +504,11 @@ func (s *Server) run(j *Job) {
req = fitted
}
}
// Under the lock: a job is read by Get while it runs, and the race
// detector caught this write beside it.
s.mu.Lock()
j.Request = req
s.mu.Unlock()
memGB := int((header.MemoryBytes(set.Bytes, req.CtxPool) + 1<<30 - 1) >> 30)
basis := fmt.Sprintf("weights %.1f GB + KV cache %.1f GB for %d tokens + engine",
gb(set.Bytes), gb(int64(req.CtxPool)*header.KVBytesPerToken()), req.CtxPool)
+14
View File
@@ -21,6 +21,7 @@ import (
"errors"
"fmt"
"log/slog"
"slices"
"sync"
"time"
@@ -248,6 +249,19 @@ func (s *Service) Resources(ctx context.Context, name, profile string, memoryGB,
return final, err
}
// SetProfiles replaces a model's profiles, as a model added with them
// would have: for a model whose engines are to run otherwise, its args among
// it. Running replicas keep what they were started with until restarted.
func (s *Service) SetProfiles(ctx context.Context, name string, profiles []catalog.Profile, by string) (catalog.Model, error) {
if len(profiles) == 0 {
return catalog.Model{}, errors.New("a model needs a profile")
}
return s.update(ctx, name, by, "replace its profiles", func(m *catalog.Model) error {
m.Profiles = slices.Clone(profiles)
return nil
})
}
// SetClass sets a model's capacity class: critical, standard or
// opportunistic, or empty to go back to what its scaling implies.
func (s *Service) SetClass(ctx context.Context, name, class, by string) (catalog.Model, error) {
+14 -1
View File
@@ -915,7 +915,20 @@ func (r *Reconciler) reconcileModel(ctx context.Context, m catalog.Model) {
// A drained replica is stopped once its replacement is serving, or once the
// grace period expires. Stopping it the moment it was drained would cut
// in-flight requests, which is the thing draining exists to avoid.
replaced := len(liveIDs) >= desired
//
// Serving counts a replica being restarted that still answers: with
// replicas restarted one after another, the second's retiring kept the
// first's drained one waiting out its grace — and with it the second's
// replacement, on a node with no room until the first had gone.
serving := len(liveIDs)
r.mu.Lock()
for _, id := range retiringIDs {
if rep, ok := r.replicas[id]; ok && rep.State == Ready {
serving++
}
}
r.mu.Unlock()
replaced := serving >= desired
for _, id := range drainingIDs {
r.mu.Lock()
rep, ok := r.replicas[id]
@@ -2436,3 +2436,51 @@ func TestTheRoomAnInPlaceRestartFreesIsItsReplacements(t *testing.T) {
}
}
}
// Two models restarted together on a node whose cores are all reserved,
// one to need more and one less: the judge fits only once the starter
// has given back its cores. The judge's restart waits for that, rather
// than being refused as if nothing would ever make room. Met on a 4-core
// single-machine install after `ctl model resources` on both.
func TestARestartWaitsForRoomAnotherRestartWillGiveBack(t *testing.T) {
cpu := inventory.Node{ID: "a", MgmtAddr: "10.0.0.1", State: inventory.Ready, MemoryTotalMB: 16 * 1024, CPUCount: 4}
judge := catalog.Model{Name: "judge", Source: "/m", Scaling: catalog.Scaling{Min: 2, Max: 2},
Profiles: []catalog.Profile{{Name: "cpu", Engine: "llamacpp", MinMemoryGB: 2, CPUCores: 1}}}
starter := catalog.Model{Name: "starter", Source: "/m", Scaling: catalog.Scaling{Min: 1, Max: 1},
Profiles: []catalog.Profile{{Name: "cpu", Engine: "llamacpp", MinMemoryGB: 2, CPUCores: 2}}}
h := newFleetHarness(t, []catalog.Model{judge, starter}, cpu)
h.settle(t, 3)
var old []string
for _, rep := range h.r.ReadyReplicas() {
old = append(old, rep.ID)
}
if len(old) != 3 {
t.Fatalf("%d replicas ready, want 3", len(old))
}
judge.Profiles[0].CPUCores, starter.Profiles[0].CPUCores = 2, 0
cat, err := catalog.New([]catalog.Model{judge, starter})
if err != nil {
t.Fatal(err)
}
h.r.cat.Store(cat)
if err := h.r.Restart(old...); err != nil {
t.Fatal(err)
}
for range 12 {
h.r.Reconcile(t.Context())
h.markReady()
h.advance(10 * time.Second)
}
for _, id := range old {
if st := h.r.RestartState(id); strings.HasPrefix(st, "refused") {
t.Errorf("%s: %s", id, st)
}
if rep, ok := replicaByID(h, id); ok && rep.State == Ready {
t.Errorf("%s still serves on its old figures", id)
}
}
if j, s := replicasOf(h, "judge"), replicasOf(h, "starter"); j != 2 || s != 1 {
t.Errorf("judge %d, starter %d after the restarts", j, s)
}
}
+28
View File
@@ -22,6 +22,10 @@ import (
// stopped now, and its replacement takes its place. Waiting would change
// nothing; it used to wait DrainGrace anyway, five minutes reading only
// "restart replacing", which looked like a restart that never would.
// - other replicas whose own restarts are under way: the replacement fits
// once they have restarted, as a starter edited to reserve fewer cores
// gives back what a judge edited to reserve more needs. Waited for, up
// to DrainGrace, so that two restarts each waiting on the other end.
// - more than it: the replacement fits nowhere even with it stopped — its
// profile was edited to need more than any node has free. Stopping it
// would leave the model with nothing, so the restart is given up, and
@@ -62,6 +66,8 @@ func (r *Reconciler) retire(ctx context.Context, m catalog.Model, ids []string,
r.log.Warn("restarting a replica in place: it is the only thing in its replacement's way",
"replica", id, "model", m.Name)
r.stopForRestart(ctx, id, "restarted in place: it was the only thing in its replacement's way")
case outcome == noRoom && r.now().Sub(since) <= r.cfg.DrainGrace && r.fitsAfterRestarts(m, id):
// Another restart under way gives the room back; waited for.
case outcome == noRoom || outcome == overLimit:
r.refuseRestart(ctx, id, "its replacement fits on no node even with this replica stopped "+
"(llmgw ctl explain-placement "+m.Name+")")
@@ -73,6 +79,28 @@ func (r *Reconciler) retire(ctx context.Context, m catalog.Model, ids []string,
}
}
// fitsAfterRestarts is whether a model's replacement for a replica fits with
// that replica gone and every other on its way out: queued to restart, being
// replaced, or draining. Their replacements take room too, so this is
// whether waiting could help, not a promise that it will.
func (r *Reconciler) fitsAfterRestarts(m catalog.Model, id string) bool {
gone := []victim{{ID: id, Model: m.Name}}
r.mu.Lock()
for oid, o := range r.replicas {
if oid == id {
continue
}
if o.State == Draining || !o.RetiringAt.IsZero() || slices.Contains(r.restartQueue[o.Model], oid) {
gone = append(gone, victim{ID: oid, Model: o.Model})
}
}
r.mu.Unlock()
if len(gone) == 1 {
return false
}
return r.fitsWithout(m, nil, gone)
}
// stopForRestart drains a replica and stops it at once, and places its
// replacement in what it freed.
//
+395
View File
@@ -0,0 +1,395 @@
// Package directory keeps accounts in step with an outside directory —
// Active Directory over LDAP — at sign-in and by a periodic sync: made for
// people the directory vouches for, given the staff role and the places in
// the tree their groups map to, and suspended when the directory disables or
// loses them.
package directory
import (
"context"
"crypto/rand"
"encoding/hex"
"errors"
"fmt"
"slices"
"strings"
"time"
"gitea.mynodes.xyz/ilker/llmgw/internal/store"
)
// Identity is a person as the directory has them.
type Identity struct {
// ID never changes for the person: an AD objectGUID.
ID string
DN string
Email string
Name string
// Groups are the person's groups, nested ones included, each by its CN
// and by its DN, so a mapping may name either.
Groups []string
// Disabled is an account the directory has turned off.
Disabled bool
}
// Mapping is what a person's groups give them: a staff role and places in
// the tree. The same shape as auth.oidc's.
type Mapping struct {
Roles map[string][]string
Memberships map[string]string
}
// Role is the highest staff role the groups give, "" when none.
func (m Mapping) Role(groups []string) string {
for _, r := range []string{store.RoleAdmin, store.RoleOperator, store.RoleBilling, store.RoleViewer} {
if slices.ContainsFunc(m.Roles[r], func(g string) bool { return hasGroup(groups, g) }) {
return r
}
}
return ""
}
// Places are the places in the tree the groups give, each once.
func (m Mapping) Places(groups []string) []string {
var out []string
for g, where := range m.Memberships {
if hasGroup(groups, g) && !slices.Contains(out, where) {
out = append(out, where)
}
}
slices.Sort(out)
return out
}
// Groups are every group the mapping names.
func (m Mapping) Groups() []string {
var out []string
add := func(g string) {
if !slices.ContainsFunc(out, func(o string) bool { return strings.EqualFold(o, g) }) {
out = append(out, g)
}
}
for _, gs := range m.Roles {
for _, g := range gs {
add(g)
}
}
for g := range m.Memberships {
add(g)
}
slices.Sort(out)
return out
}
// hasGroup matches a group's CN or DN as AD does: case aside.
func hasGroup(groups []string, g string) bool {
return slices.ContainsFunc(groups, func(have string) bool { return strings.EqualFold(have, g) })
}
// Store is what keeping accounts needs of the store.
type Store interface {
UserByEmail(ctx context.Context, email string) (store.User, error)
UserByID(ctx context.Context, id string) (store.User, error)
CreateUser(ctx context.Context, u store.User) error
SetStaffRole(ctx context.Context, id, role string) error
SetUserState(ctx context.Context, id string, state store.UserState) error
DeleteUserSessions(ctx context.Context, userID string) error
AddMembership(ctx context.Context, user string, scope store.Scope, id, role, by string, at time.Time) error
RemoveMembership(ctx context.Context, userID string, scope store.Scope, scopeID string, at time.Time) (int, error)
MembershipsOf(ctx context.Context, userID string) ([]store.Membership, error)
AllOrgs(ctx context.Context) ([]store.Org, error)
AllTeams(ctx context.Context) ([]store.TeamPath, error)
AllProjects(ctx context.Context) ([]store.ProjectPath, error)
DirectoryUser(ctx context.Context, directory, externalID string) (store.DirectoryLink, error)
LinkDirectory(ctx context.Context, userID, directory, externalID string, at time.Time) error
DirectoryLinks(ctx context.Context, directory string) ([]store.DirectoryLink, error)
SetSyncSuspended(ctx context.Context, userID string, suspended bool, at time.Time) error
}
// Keeper keeps one directory's accounts.
type Keeper struct {
St Store
// Name is the directory's, as links and memberships record it: "ldap".
Name string
Mapping Mapping
CreateAccounts bool
Now func() time.Time
}
// The ways a person the directory vouches for is still refused.
var (
ErrNoAccount = errors.New("no account here, and accounts are not made for the directory's people")
ErrSuspended = errors.New("the account is suspended here")
ErrNotActive = errors.New("the account is not active here")
ErrNoEmail = errors.New("the directory has no mail for this person")
// ErrImplausible is a sync that would suspend too many at once: done
// nothing, and said.
ErrImplausible = errors.New("the directory's answer is implausible; nobody was suspended")
)
func (k *Keeper) now() time.Time {
if k.Now != nil {
return k.Now()
}
return time.Now()
}
// SignIn is the account for a person the directory has just vouched for:
// found by their directory id, else their mail, else made; given what their
// groups map to; refused when it may not sign in.
func (k *Keeper) SignIn(ctx context.Context, id Identity) (store.User, error) {
u, link, err := k.account(ctx, id)
if err != nil {
return store.User{}, err
}
// The directory has them enabled, or the bind would have failed: an
// account the sync suspended for being disabled there is theirs again.
if u.State == store.UserSuspended && link.SuspendedBySync {
if u, err = k.reinstate(ctx, u); err != nil {
return store.User{}, err
}
}
switch {
case u.State == store.UserSuspended:
return store.User{}, ErrSuspended
case !u.CanSignIn():
return store.User{}, fmt.Errorf("%w: %s", ErrNotActive, u.State)
}
if err := k.apply(ctx, &u, id, false); err != nil {
return store.User{}, err
}
return u, nil
}
// account finds or makes the account and links it to the directory.
func (k *Keeper) account(ctx context.Context, id Identity) (store.User, store.DirectoryLink, error) {
link, err := k.St.DirectoryUser(ctx, k.Name, id.ID)
switch {
case err == nil:
u, err := k.St.UserByID(ctx, link.UserID)
return u, link, err
case !errors.Is(err, store.ErrNotFound):
return store.User{}, link, err
}
email := store.NormalizeEmail(id.Email)
if email == "" {
return store.User{}, link, ErrNoEmail
}
u, err := k.St.UserByEmail(ctx, email)
switch {
case errors.Is(err, store.ErrNotFound) && k.CreateAccounts:
name := id.Name
if name == "" {
name = email
}
u = store.User{ID: "usr-" + randomID(), Email: email, Name: name,
// No password works here: the directory's is checked.
PasswordHash: "!" + k.Name, State: store.UserActive, CreatedAt: k.now()}
if err := k.St.CreateUser(ctx, u); err != nil {
return store.User{}, link, err
}
case errors.Is(err, store.ErrNotFound):
return store.User{}, link, fmt.Errorf("%w: %s", ErrNoAccount, email)
case err != nil:
return store.User{}, link, err
}
if err := k.St.LinkDirectory(ctx, u.ID, k.Name, id.ID, k.now()); err != nil {
return store.User{}, link, err
}
return u, store.DirectoryLink{UserID: u.ID, Directory: k.Name, ExternalID: id.ID}, nil
}
// apply gives the account the role and places its groups map to. With
// remove, a place this directory gave that the groups no longer do is taken
// away; one given otherwise is never touched.
func (k *Keeper) apply(ctx context.Context, u *store.User, id Identity, remove bool) error {
if len(k.Mapping.Roles) > 0 {
if role := k.Mapping.Role(id.Groups); role != u.StaffRole() {
if err := k.St.SetStaffRole(ctx, u.ID, role); err != nil {
return err
}
u.IsAdmin, u.Role = role == store.RoleAdmin, role
if role == store.RoleAdmin {
u.Role = ""
}
}
}
want := map[string]store.Scope{}
for _, where := range k.Mapping.Places(id.Groups) {
scope, sid, err := k.place(ctx, where)
if err != nil {
continue // a mapping to a place the tree lacks: said by Sync
}
want[string(scope)+"/"+sid] = scope
if err := k.St.AddMembership(ctx, u.ID, scope, sid, store.RoleMember, k.Name, k.now()); err != nil {
return err
}
}
if !remove {
return nil
}
have, err := k.St.MembershipsOf(ctx, u.ID)
if err != nil {
return err
}
for _, m := range have {
if m.AddedBy != k.Name {
continue
}
if _, ok := want[string(m.Scope)+"/"+m.ScopeID]; ok {
continue
}
if _, err := k.St.RemoveMembership(ctx, u.ID, m.Scope, m.ScopeID, k.now()); err != nil {
return err
}
}
return nil
}
func (k *Keeper) reinstate(ctx context.Context, u store.User) (store.User, error) {
if err := k.St.SetUserState(ctx, u.ID, store.UserActive); err != nil {
return u, err
}
if err := k.St.SetSyncSuspended(ctx, u.ID, false, k.now()); err != nil {
return u, err
}
u.State = store.UserActive
return u, nil
}
// SyncReport is what a sync did.
type SyncReport struct {
Seen, Updated, Suspended, Reinstated int
// Unmapped are mappings to places the tree does not have.
Unmapped []string
}
// Sync brings every account the directory keeps in step with what it says
// now: people disabled or gone there are suspended here (their sessions
// ended; their keys refuse, as a suspended owner's do), those enabled again
// that the sync had suspended are reinstated, and everyone's role and places
// follow their groups. People the directory has that never signed in here
// are left until they do.
func (k *Keeper) Sync(ctx context.Context, people []Identity) (SyncReport, error) {
var rep SyncReport
rep.Unmapped = k.unmapped(ctx)
byID := make(map[string]Identity, len(people))
for _, p := range people {
byID[p.ID] = p
}
links, err := k.St.DirectoryLinks(ctx, k.Name)
if err != nil {
return rep, err
}
// A directory that answers with nobody, or with half the people missing,
// is far likelier a wrong base, a filter or a bind account's lost rights
// than half the company leaving at once. Nothing is suspended for it.
gone := 0
for _, l := range links {
if p, there := byID[l.ExternalID]; !there || p.Disabled {
gone++
}
}
if len(links) > 0 && (len(people) == 0 || (gone >= 3 && gone*2 > len(links))) {
return rep, fmt.Errorf("%w: %d of the %d accounts it keeps would be suspended", ErrImplausible, gone, len(links))
}
for _, l := range links {
rep.Seen++
u, err := k.St.UserByID(ctx, l.UserID)
if err != nil {
continue
}
p, there := byID[l.ExternalID]
switch {
case !there || p.Disabled:
if u.State == store.UserActive {
if err := k.St.SetUserState(ctx, u.ID, store.UserSuspended); err != nil {
return rep, err
}
if err := k.St.DeleteUserSessions(ctx, u.ID); err != nil {
return rep, err
}
if err := k.St.SetSyncSuspended(ctx, u.ID, true, k.now()); err != nil {
return rep, err
}
rep.Suspended++
}
default:
if u.State == store.UserSuspended && l.SuspendedBySync {
if u, err = k.reinstate(ctx, u); err != nil {
return rep, err
}
rep.Reinstated++
}
if u.State != store.UserActive {
continue
}
before := u.StaffRole()
if err := k.apply(ctx, &u, p, true); err != nil {
return rep, err
}
if u.StaffRole() != before {
rep.Updated++
}
}
}
return rep, nil
}
// unmapped are the mapping's places the tree does not have.
func (k *Keeper) unmapped(ctx context.Context) []string {
var out []string
for _, where := range k.Mapping.Memberships {
if _, _, err := k.place(ctx, where); err != nil && !slices.Contains(out, where) {
out = append(out, where)
}
}
slices.Sort(out)
return out
}
// place finds org:NAME, team:ORG/TEAM, project:ORG/TEAM/PROJECT, as
// auth.oidc's mapping does.
func (k *Keeper) place(ctx context.Context, where string) (store.Scope, string, error) {
kind, path, _ := strings.Cut(where, ":")
norm := func(s string) string {
parts := strings.Split(s, "/")
for i := range parts {
parts[i] = strings.TrimSpace(parts[i])
}
return strings.Join(parts, "/")
}
same := func(a string) bool { return strings.EqualFold(norm(a), norm(path)) }
switch kind {
case "org":
orgs, err := k.St.AllOrgs(ctx)
for _, o := range orgs {
if same(o.Name) {
return store.ScopeOrg, o.ID, nil
}
}
return "", "", errors.Join(err, errors.New("no such organization"))
case "team":
teams, err := k.St.AllTeams(ctx)
for _, t := range teams {
if same(t.OrgName + "/" + t.Name) {
return store.ScopeTeam, t.ID, nil
}
}
return "", "", errors.Join(err, errors.New("no such team"))
default:
projects, err := k.St.AllProjects(ctx)
for _, p := range projects {
if same(p.Path) {
return store.ScopeProject, p.ID, nil
}
}
return "", "", errors.Join(err, errors.New("no such project"))
}
}
func randomID() string {
var b [8]byte
_, _ = rand.Read(b[:])
return hex.EncodeToString(b[:])
}
+142
View File
@@ -0,0 +1,142 @@
package directory
import (
"errors"
"path/filepath"
"slices"
"testing"
"time"
"gitea.mynodes.xyz/ilker/llmgw/internal/store"
)
func newKeeper(t *testing.T) (*Keeper, *store.DB, string) {
t.Helper()
st, err := store.Open(t.Context(), filepath.Join(t.TempDir(), "s.db"))
if err != nil {
t.Fatal(err)
}
t.Cleanup(func() { st.Close() })
org, _ := st.CreateOrg(t.Context(), "Acme")
team, _ := st.CreateTeam(t.Context(), org, "ml")
k := &Keeper{St: st, Name: "ldap", CreateAccounts: true, Mapping: Mapping{
Roles: map[string][]string{"admin": {"llmgw-admins"}, "viewer": {"CN=Staff,OU=Groups,DC=corp,DC=example"}},
Memberships: map[string]string{"ml-team": "team:Acme/ml", "ghosts": "team:Acme/nowhere"},
}}
return k, st, team
}
var ada = Identity{ID: "guid-ada", DN: "CN=Ada,OU=Staff,DC=corp,DC=example", Email: "Ada@Corp.example", Name: "Ada L",
Groups: []string{"ml-team", "CN=ml-team,OU=Groups,DC=corp,DC=example", "Staff", "CN=Staff,OU=Groups,DC=corp,DC=example"}}
func TestAPersonTheDirectoryVouchesForIsMadeAnAccountWithWhatTheirGroupsGive(t *testing.T) {
k, st, team := newKeeper(t)
u, err := k.SignIn(t.Context(), ada)
if err != nil {
t.Fatal(err)
}
if u.Email != "ada@corp.example" || u.Name != "Ada L" || u.PasswordHash != "!ldap" || u.StaffRole() != "viewer" {
t.Errorf("account %+v role %q", u, u.StaffRole())
}
ms, _ := st.MembershipsOf(t.Context(), u.ID)
if len(ms) != 1 || ms[0].ScopeID != team || ms[0].AddedBy != "ldap" {
t.Errorf("memberships %+v", ms)
}
// Their mail changes there: still the same account, by its directory id.
moved := ada
moved.Email = "ada.lovelace@corp.example"
again, err := k.SignIn(t.Context(), moved)
if err != nil || again.ID != u.ID {
t.Errorf("a new mail made a new account: %v %+v", err, again)
}
// The higher role of two wins.
boss := ada
boss.Groups = append(slices.Clone(ada.Groups), "llmgw-admins")
if u, _ := k.SignIn(t.Context(), boss); u.StaffRole() != "admin" {
t.Errorf("role %q", u.StaffRole())
}
}
func TestWithoutAccountsMadeOnlyThoseThatExistSignIn(t *testing.T) {
k, _, _ := newKeeper(t)
k.CreateAccounts = false
if _, err := k.SignIn(t.Context(), ada); !errors.Is(err, ErrNoAccount) {
t.Errorf("err %v", err)
}
}
func TestTheSyncSuspendsWhomTheDirectoryDisabledAndReinstatesThemAfter(t *testing.T) {
k, st, _ := newKeeper(t)
ctx := t.Context()
people := []Identity{ada}
for _, n := range []string{"b", "c", "d"} {
people = append(people, Identity{ID: "guid-" + n, Email: n + "@corp.example"})
}
ids := map[string]string{}
for _, p := range people {
u, err := k.SignIn(ctx, p)
if err != nil {
t.Fatal(err)
}
ids[p.ID] = u.ID
}
if err := st.CreateSession(ctx, ids["guid-ada"], "tok", "csrf", time.Hour); err != nil {
t.Fatal(err)
}
// An administrator suspends c here, by hand.
_ = st.SetUserState(ctx, ids["guid-c"], store.UserSuspended)
// Ada is disabled there, and has left ml-team.
off := ada
off.Disabled = true
rep, err := k.Sync(ctx, []Identity{off, people[1], people[2], people[3]})
if err != nil {
t.Fatal(err)
}
u, _ := st.UserByID(ctx, ids["guid-ada"])
if u.State != store.UserSuspended || rep.Suspended != 1 {
t.Errorf("ada %s, report %+v", u.State, rep)
}
if _, _, err := st.SessionByToken(ctx, "tok"); err == nil {
t.Error("a suspended account's session still works")
}
if !slices.Contains(rep.Unmapped, "team:Acme/nowhere") {
t.Errorf("an unmapped place was not said: %+v", rep)
}
// Enabled again, and out of ml-team: reinstated, and its place taken
// away. c, suspended by hand, stays so.
back := ada
back.Groups = []string{"Staff", "CN=Staff,OU=Groups,DC=corp,DC=example"}
rep, err = k.Sync(ctx, []Identity{back, people[1], people[2], people[3]})
if err != nil {
t.Fatal(err)
}
u, _ = st.UserByID(ctx, ids["guid-ada"])
ms, _ := st.MembershipsOf(ctx, u.ID)
if u.State != store.UserActive || rep.Reinstated != 1 || len(ms) != 0 {
t.Errorf("ada %s, memberships %+v, report %+v", u.State, ms, rep)
}
if c, _ := st.UserByID(ctx, ids["guid-c"]); c.State != store.UserSuspended {
t.Errorf("an account suspended by hand was reinstated: %s", c.State)
}
}
func TestASyncThatWouldSuspendMostIsRefused(t *testing.T) {
k, st, _ := newKeeper(t)
ctx := t.Context()
for _, n := range []string{"a", "b", "c", "d"} {
if _, err := k.SignIn(ctx, Identity{ID: "guid-" + n, Email: n + "@corp.example"}); err != nil {
t.Fatal(err)
}
}
for _, people := range [][]Identity{nil, {{ID: "guid-a", Email: "a@corp.example"}}} {
if _, err := k.Sync(ctx, people); !errors.Is(err, ErrImplausible) {
t.Errorf("%d people: %v", len(people), err)
}
}
users, _ := st.Users(ctx, store.UserSuspended)
if len(users) != 0 {
t.Errorf("suspended %d", len(users))
}
}
+286
View File
@@ -0,0 +1,286 @@
package directory
import (
"context"
"crypto/tls"
"crypto/x509"
"encoding/binary"
"errors"
"fmt"
"os"
"strconv"
"strings"
"github.com/go-ldap/ldap/v3"
"gitea.mynodes.xyz/ilker/llmgw/internal/config"
"gitea.mynodes.xyz/ilker/llmgw/internal/secrets"
)
// LDAP is a directory reached over LDAP, written for Active Directory: its
// attribute names, its disabled flag, and its matching rule for nested
// groups (LDAP_MATCHING_RULE_IN_CHAIN).
type LDAP struct {
cfg config.LDAP
tls *tls.Config
dial func(url string) (ldapConn, error)
}
// ldapConn is the part of a connection used here, so tests need no server.
type ldapConn interface {
Bind(user, password string) error
Search(*ldap.SearchRequest) (*ldap.SearchResult, error)
SearchWithPaging(*ldap.SearchRequest, uint32) (*ldap.SearchResult, error)
Close() error
}
// inChain is AD's matching rule that follows nested group membership.
const inChain = "1.2.840.113556.1.4.1941"
// uacDisabled is userAccountControl's ACCOUNTDISABLE bit.
const uacDisabled = 0x2
var personAttrs = []string{"objectGUID", "sAMAccountName", "userPrincipalName", "mail", "displayName", "userAccountControl"}
// ErrBadCredentials is a person the directory does not know, or a password
// it does not take: said alike, so a form cannot tell which.
var ErrBadCredentials = errors.New("the directory does not take that name and password")
// NewLDAP prepares a directory from its configuration.
func NewLDAP(cfg config.LDAP) (*LDAP, error) {
tc := &tls.Config{MinVersion: tls.VersionTLS12}
if cfg.CAFile != "" {
pem, err := os.ReadFile(cfg.CAFile)
if err != nil {
return nil, fmt.Errorf("auth.ldap.ca_file: %w", err)
}
pool := x509.NewCertPool()
if !pool.AppendCertsFromPEM(pem) {
return nil, fmt.Errorf("auth.ldap.ca_file %s holds no certificate", cfg.CAFile)
}
tc.RootCAs = pool
}
d := &LDAP{cfg: cfg, tls: tc}
d.dial = d.dialURL
return d, nil
}
func (d *LDAP) dialURL(u string) (ldapConn, error) {
c, err := ldap.DialURL(u, ldap.DialWithTLSConfig(d.tls))
if err != nil {
return nil, err
}
c.SetTimeout(d.cfg.RoundTrip())
if strings.HasPrefix(u, "ldap://") {
// Validate refuses ldap:// without start_tls: nothing is sent first.
if err := c.StartTLS(d.tls); err != nil {
c.Close()
return nil, fmt.Errorf("StartTLS: %w", err)
}
}
return c, nil
}
// service is a connection bound as the account that finds people, to the
// first server that answers.
func (d *LDAP) service(ctx context.Context) (ldapConn, error) {
pw, err := secrets.Read(ctx, d.cfg.BindPasswordFile)
if err != nil {
return nil, fmt.Errorf("auth.ldap.bind_password_file: %w", err)
}
var errs []error
for _, u := range d.cfg.URLs {
c, err := d.dial(u)
if err != nil {
errs = append(errs, fmt.Errorf("%s: %w", u, err))
continue
}
if err := c.Bind(d.cfg.BindDN, strings.TrimSpace(string(pw))); err != nil {
c.Close()
// A server that refuses the service account refuses it
// everywhere: not tried on the next.
return nil, fmt.Errorf("binding as %s on %s: %w", d.cfg.BindDN, u, err)
}
return c, nil
}
return nil, fmt.Errorf("no directory server answered: %w", errors.Join(errs...))
}
// Authenticate checks a person's name and password against the directory
// and says who they are there, groups and all.
func (d *LDAP) Authenticate(ctx context.Context, login, password string) (Identity, error) {
login = strings.TrimSpace(login)
// An empty password is an unauthenticated bind, which succeeds: never
// one for a person.
if login == "" || password == "" {
return Identity{}, ErrBadCredentials
}
// DOMAIN\user, as Windows writes it.
if i := strings.LastIndexByte(login, '\\'); i >= 0 {
login = login[i+1:]
}
c, err := d.service(ctx)
if err != nil {
return Identity{}, err
}
defer c.Close()
q := ldap.EscapeFilter(login)
filter := "(&" + d.people(false) + "(|(sAMAccountName=" + q + ")(userPrincipalName=" + q + ")(mail=" + q + ")))"
res, err := c.Search(ldap.NewSearchRequest(d.cfg.BaseDN, ldap.ScopeWholeSubtree, ldap.NeverDerefAliases,
2, 0, false, filter, personAttrs, nil))
if err != nil {
return Identity{}, fmt.Errorf("finding %s: %w", login, err)
}
if len(res.Entries) != 1 {
// None, or two people one name matches: neither is let in.
return Identity{}, ErrBadCredentials
}
id := identity(res.Entries[0])
if err := c.Bind(id.DN, password); err != nil {
var le *ldap.Error
if errors.As(err, &le) && le.ResultCode == ldap.LDAPResultInvalidCredentials {
return Identity{}, ErrBadCredentials
}
return Identity{}, fmt.Errorf("checking %s's password: %w", login, err)
}
// Back to the service account to read groups, which a person may not.
pw, err := secrets.Read(ctx, d.cfg.BindPasswordFile)
if err != nil {
return Identity{}, err
}
if err := c.Bind(d.cfg.BindDN, strings.TrimSpace(string(pw))); err != nil {
return Identity{}, err
}
if id.Groups, err = d.groupsOf(c, id.DN); err != nil {
return Identity{}, err
}
return id, nil
}
// groupsOf are a person's groups, nested ones included, by CN and DN.
func (d *LDAP) groupsOf(c ldapConn, dn string) ([]string, error) {
filter := "(&(objectClass=group)(member:" + inChain + ":=" + ldap.EscapeFilter(dn) + "))"
res, err := c.SearchWithPaging(ldap.NewSearchRequest(d.groupBase(), ldap.ScopeWholeSubtree, ldap.NeverDerefAliases,
0, 0, false, filter, []string{"cn"}, nil), 500)
if err != nil {
return nil, fmt.Errorf("reading the groups of %s: %w", dn, err)
}
var out []string
for _, e := range res.Entries {
out = append(out, e.GetAttributeValue("cn"), e.DN)
}
return out, nil
}
// People are everyone the directory has under the base, enabled or not,
// each with the mapped groups they are in: what a sync compares against.
func (d *LDAP) People(ctx context.Context, m Mapping) ([]Identity, error) {
c, err := d.service(ctx)
if err != nil {
return nil, err
}
defer c.Close()
res, err := c.SearchWithPaging(ldap.NewSearchRequest(d.cfg.BaseDN, ldap.ScopeWholeSubtree, ldap.NeverDerefAliases,
0, 0, false, d.people(true), personAttrs, nil), 500)
if err != nil {
return nil, fmt.Errorf("listing people: %w", err)
}
people := make([]Identity, 0, len(res.Entries))
at := map[string]int{}
for _, e := range res.Entries {
id := identity(e)
at[id.ID] = len(people)
people = append(people, id)
}
// Only the groups the mapping names matter: a query each, for its
// members nested ones included, rather than one per person.
for _, g := range m.Groups() {
dns, err := d.groupDNs(c, g)
if err != nil {
return nil, err
}
for _, gdn := range dns {
filter := "(&" + d.people(true) + "(memberOf:" + inChain + ":=" + ldap.EscapeFilter(gdn) + "))"
r, err := c.SearchWithPaging(ldap.NewSearchRequest(d.cfg.BaseDN, ldap.ScopeWholeSubtree,
ldap.NeverDerefAliases, 0, 0, false, filter, []string{"objectGUID"}, nil), 500)
if err != nil {
return nil, fmt.Errorf("reading the members of %s: %w", gdn, err)
}
for _, e := range r.Entries {
if i, ok := at[guid(e.GetRawAttributeValue("objectGUID"))]; ok {
people[i].Groups = append(people[i].Groups, g, gdn)
}
}
}
}
return people, nil
}
// groupDNs are the groups a mapping's name means: a DN as it is, a CN
// looked up.
func (d *LDAP) groupDNs(c ldapConn, g string) ([]string, error) {
if strings.Contains(g, "=") {
return []string{g}, nil
}
res, err := c.Search(ldap.NewSearchRequest(d.groupBase(), ldap.ScopeWholeSubtree, ldap.NeverDerefAliases,
0, 0, false, "(&(objectClass=group)(cn="+ldap.EscapeFilter(g)+"))", []string{"cn"}, nil))
if err != nil {
return nil, fmt.Errorf("finding the group %s: %w", g, err)
}
var out []string
for _, e := range res.Entries {
out = append(out, e.DN)
}
return out, nil
}
// people is the filter for person accounts under the user filter; enabled
// ones only, unless all.
func (d *LDAP) people(all bool) string {
f := "(objectCategory=person)(objectClass=user)"
if !all {
f += "(!(userAccountControl:1.2.840.113556.1.4.803:=" + strconv.Itoa(uacDisabled) + "))"
}
return "(&" + f + d.cfg.UserFilter + ")"
}
// groupBase is where groups are searched: GroupBaseDN, else the domain root
// of BaseDN (its DC= parts).
func (d *LDAP) groupBase() string {
if d.cfg.GroupBaseDN != "" {
return d.cfg.GroupBaseDN
}
var dc []string
for _, rdn := range strings.Split(d.cfg.BaseDN, ",") {
if strings.HasPrefix(strings.ToUpper(strings.TrimSpace(rdn)), "DC=") {
dc = append(dc, strings.TrimSpace(rdn))
}
}
if len(dc) == 0 {
return d.cfg.BaseDN
}
return strings.Join(dc, ",")
}
func identity(e *ldap.Entry) Identity {
email := e.GetAttributeValue("mail")
if email == "" {
email = e.GetAttributeValue("userPrincipalName")
}
uac, _ := strconv.Atoi(e.GetAttributeValue("userAccountControl"))
return Identity{
ID: guid(e.GetRawAttributeValue("objectGUID")), DN: e.DN,
Email: email, Name: e.GetAttributeValue("displayName"),
Disabled: uac&uacDisabled != 0,
}
}
// guid is an objectGUID as Windows writes it: its first three parts
// little-endian. Another directory's id is kept as hex.
func guid(b []byte) string {
if len(b) != 16 {
return fmt.Sprintf("%x", b)
}
return fmt.Sprintf("%08x-%04x-%04x-%x-%x", binary.LittleEndian.Uint32(b[0:4]),
binary.LittleEndian.Uint16(b[4:6]), binary.LittleEndian.Uint16(b[6:8]), b[8:10], b[10:16])
}
+130
View File
@@ -0,0 +1,130 @@
package directory
import (
"errors"
"os"
"path/filepath"
"strings"
"testing"
"github.com/go-ldap/ldap/v3"
"gitea.mynodes.xyz/ilker/llmgw/internal/config"
)
// fakeConn is a directory with one person in it.
type fakeConn struct {
binds []string
filters []string
password string // the person's
people int // how many the person search finds
}
func (f *fakeConn) Bind(user, password string) error {
f.binds = append(f.binds, user)
if strings.HasPrefix(user, "CN=Ada") && password != f.password {
return &ldap.Error{ResultCode: ldap.LDAPResultInvalidCredentials}
}
return nil
}
func (f *fakeConn) Search(r *ldap.SearchRequest) (*ldap.SearchResult, error) {
f.filters = append(f.filters, r.Filter)
var out []*ldap.Entry
for i := range f.people {
e := ldap.NewEntry("CN=Ada,OU=Staff,DC=corp,DC=example", map[string][]string{
"mail": {"ada@corp.example"}, "displayName": {"Ada"}, "userAccountControl": {"512"},
})
e.Attributes = append(e.Attributes, &ldap.EntryAttribute{Name: "objectGUID", ByteValues: [][]byte{guidBytes(byte(i))}})
out = append(out, e)
}
return &ldap.SearchResult{Entries: out}, nil
}
func (f *fakeConn) SearchWithPaging(r *ldap.SearchRequest, _ uint32) (*ldap.SearchResult, error) {
f.filters = append(f.filters, r.Filter)
return &ldap.SearchResult{Entries: []*ldap.Entry{
ldap.NewEntry("CN=ml-team,OU=Groups,DC=corp,DC=example", map[string][]string{"cn": {"ml-team"}}),
}}, nil
}
func (f *fakeConn) Close() error { return nil }
func guidBytes(last byte) []byte {
return []byte{0x78, 0x56, 0x34, 0x12, 0x34, 0x12, 0x78, 0x56, 0x9a, 0xbc, 0xde, 0xf0, 0x12, 0x34, 0x56, last}
}
func fakeLDAP(t *testing.T, c *fakeConn) *LDAP {
t.Helper()
pw := filepath.Join(t.TempDir(), "bind")
_ = os.WriteFile(pw, []byte("svc-secret\n"), 0o600)
d, err := NewLDAP(config.LDAP{URLs: []string{"ldaps://dc1"}, BindDN: "CN=llmgw,OU=Svc,DC=corp,DC=example",
BindPasswordFile: pw, BaseDN: "OU=Staff,DC=corp,DC=example", UserFilter: "(memberOf=CN=llm,DC=corp,DC=example)"})
if err != nil {
t.Fatal(err)
}
d.dial = func(string) (ldapConn, error) { return c, nil }
return d
}
func TestAPersonIsCheckedByBindingAsThemAndTheirGroupsReadNested(t *testing.T) {
c := &fakeConn{password: "ad-secret", people: 1}
d := fakeLDAP(t, c)
id, err := d.Authenticate(t.Context(), `CORP\ada*)(x=`, "ad-secret")
if err != nil {
t.Fatal(err)
}
if id.Email != "ada@corp.example" || id.ID != "12345678-1234-5678-9abc-def012345600" ||
!slicesHas(id.Groups, "ml-team") || !slicesHas(id.Groups, "CN=ml-team,OU=Groups,DC=corp,DC=example") {
t.Errorf("identity %+v", id)
}
// The name is escaped into the filter, the domain part dropped; the
// user filter is ANDed in; disabled accounts are not found; groups are
// read with the in-chain rule.
find, groups := c.filters[0], c.filters[1]
if !strings.Contains(find, `(sAMAccountName=ada\2a\29\28x=)`) || !strings.Contains(find, "(memberOf=CN=llm,DC=corp,DC=example)") ||
!strings.Contains(find, "1.2.840.113556.1.4.803:=2") || !strings.Contains(groups, inChain) {
t.Errorf("filters %q", c.filters)
}
// Service account, then the person, then the service account again.
if len(c.binds) != 3 || c.binds[1] != "CN=Ada,OU=Staff,DC=corp,DC=example" {
t.Errorf("binds %v", c.binds)
}
}
func TestWrongPasswordsEmptyOnesAndAmbiguousNamesAreRefusedAlike(t *testing.T) {
for _, tc := range []struct {
people int
password string
}{{1, "wrong"}, {1, ""}, {0, "ad-secret"}, {2, "ad-secret"}} {
d := fakeLDAP(t, &fakeConn{password: "ad-secret", people: tc.people})
if _, err := d.Authenticate(t.Context(), "ada", tc.password); !errors.Is(err, ErrBadCredentials) {
t.Errorf("%d found, password %q: %v", tc.people, tc.password, err)
}
}
}
func TestPlaintextLDAPIsRefusedWithoutStartTLS(t *testing.T) {
base := config.LDAP{BindDN: "CN=x", BindPasswordFile: "/f", BaseDN: "DC=corp"}
for _, c := range []struct {
urls []string
start bool
ok bool
}{{[]string{"ldaps://dc1"}, false, true}, {[]string{"ldap://dc1"}, false, false},
{[]string{"ldap://dc1"}, true, true}, {[]string{"http://dc1"}, true, false}} {
l := base
l.URLs, l.StartTLS = c.urls, c.start
if err := l.Validate(); (err == nil) != c.ok {
t.Errorf("%v start_tls=%v: %v", c.urls, c.start, err)
}
}
}
func slicesHas(xs []string, x string) bool {
for _, v := range xs {
if v == x {
return true
}
}
return false
}
+22
View File
@@ -33,6 +33,7 @@ var Reference = []Block{
{Key: "default_tpm", Type: "int", Doc: "", Nested: ""},
{Key: "signup_open", Type: "bool", Doc: "SignupOpen serves the signup form. Accounts it creates are pending until\nan operator approves them, so this is not a way in — but turning it off\nmeans accounts are made with `llmgw users create` instead.", Nested: ""},
{Key: "oidc", Type: "OIDC", Doc: "OIDC is sign-in through an identity provider.", Nested: "config.OIDC"},
{Key: "ldap", Type: "LDAP", Doc: "LDAP is sign-in against a directory, Active Directory above all, and\nthe accounts kept in step with it.", Nested: "config.LDAP"},
{Key: "scim", Type: "SCIM", Doc: "SCIM is account provisioning by the identity provider.", Nested: "config.SCIM"},
{Key: "max_key_ttl", Type: "time.Duration", Doc: "MaxKeyTTL is how long any API key may live; a key asked for with no\nexpiry is given this. Organizations, teams and projects can set\nshorter ones (llmgw keys ttl). Zero is no limit.", Nested: ""},
{Key: "personal", Type: "Personal", Doc: "Personal is what a person may do on their own, outside any project: make\npersonal keys and use the chat page personally, within these limits.", Nested: "config.Personal"},
@@ -125,11 +126,16 @@ var Reference = []Block{
{Key: "url", Type: "string", Doc: "URL is the provider's base address, without /v1: https://api.openai.com.", Nested: ""},
{Key: "model", Type: "string", Doc: "Model is what the provider calls it; requests are rewritten to it.", Nested: ""},
{Key: "api_key_file", Type: "string", Doc: "APIKeyFile holds the provider's key, on the control plane. A file\nrather than the key itself, so the key is in neither the catalog, the\nstore nor the audit log; empty for a provider that asks for none.", Nested: ""},
{Key: "api_key_files", Type: "[]ExternalKey", Doc: "APIKeyFiles are several keys for the provider, in place of\nAPIKeyFile: one endpoint each (external:MODEL#1, #2…), so a key that\nanswers 429 or 401 is passed over while its breaker is open and the\nrequest goes to another. A provider's limits are per key.", Nested: ""},
{Key: "peer", Type: "string", Doc: "Peer names the federated llmgw cluster this model is offered by, for\na model that federation added (`llmgw ctl federation add`) rather\nthan one written by hand. Such a model is named PEER/MODEL, comes and\ngoes with the peer's own list, and is sent only while the peer is up.", Nested: ""},
{Key: "api", Type: "string", Doc: "API is what the provider speaks: OpenAI's, unless this says comfyui —\na ComfyUI server, asked for pictures at /v1/images/generations and\nsent the workflow below with the request's prompt in it. Model is then\nthe checkpoint file the workflow loads, as ComfyUI lists it.", Nested: ""},
{Key: "workflow", Type: "string", Doc: "Workflow is the ComfyUI workflow, in its API format: a built-in's name\n(flux-schnell, the default, or flux-dev) or a JSON file on the control\nplane.", Nested: ""},
{Key: "output", Type: "string", Doc: "Output is what it makes: pictures, or meshes (see ComfyUI.Output).", Nested: ""},
}},
{Package: "catalog", Name: "ExternalKey", Title: "ExternalKey", Doc: "", Fields: []Field{
{Key: "file", Type: "string", Doc: "File holds the key, on the control plane, as APIKeyFile does.", Nested: ""},
{Key: "weight", Type: "float64", Doc: "Weight is its share of the requests against the others'; zero is 1.", Nested: ""},
}},
{Package: "config", Name: "Facility", Title: "Facility", Doc: "", Fields: []Field{
{Key: "budgets", Type: "[]FacilityBudget", Doc: "Budgets are the domains with a power cap.", Nested: ""},
{Key: "margin", Type: "float64", Doc: "Margin is the share of every cap kept free, 0.1 for ten percent. Zero\nkeeps none, for caps already given derated.", Nested: ""},
@@ -183,6 +189,22 @@ var Reference = []Block{
{Key: "min_blocks", Type: "int", Doc: "MinBlocks is the fewest leading blocks a replica must hold to be\npreferred for them; fewer is coincidence — a chat template's opening\ntokens — rather than a prefix worth following. Zero takes 4, which is\n64 tokens at vLLM's usual block size of 16.", Nested: ""},
{Key: "max_blocks", Type: "int", Doc: "MaxBlocks bounds how many of a prompt's blocks are hashed and looked\nup. Beyond it a longer match still helps, but the difference no longer\ndecides anything. Zero takes 4096.", Nested: ""},
}},
{Package: "config", Name: "LDAP", Title: "LDAP", Doc: "", Fields: []Field{
{Key: "urls", Type: "[]string", Doc: "URLs are the directory's servers, tried in turn: ldaps:// (636), or\nldap:// (389) with StartTLS.", Nested: ""},
{Key: "start_tls", Type: "bool", Doc: "StartTLS upgrades an ldap:// connection before anything is sent. A\npassword never crosses a connection without TLS: ldap:// without it\nis refused.", Nested: ""},
{Key: "ca_file", Type: "string", Doc: "CAFile is the authority the servers' certificates are checked against,\nwhen it is not the system's: an AD Certificate Services root.", Nested: ""},
{Key: "bind_dn", Type: "string", Doc: "BindDN and the file holding its password: a service account that may\nread users and groups. Nothing is written to the directory.", Nested: ""},
{Key: "bind_password_file", Type: "string", Doc: "", Nested: ""},
{Key: "base_dn", Type: "string", Doc: "BaseDN is where people are searched for, subtree included.", Nested: ""},
{Key: "group_base_dn", Type: "string", Doc: "GroupBaseDN is where their groups are; empty is BaseDN's domain root.", Nested: ""},
{Key: "user_filter", Type: "string", Doc: "UserFilter narrows who may sign in, ANDed with the login match: a\ngroup (memberOf=...) or an OU's people only. Empty is every enabled\nuser object.", Nested: ""},
{Key: "label", Type: "string", Doc: "Label is what the form says to sign in with: \"Sign in with your\n<label> account\". Empty is \"Active Directory\".", Nested: ""},
{Key: "create_accounts", Type: "bool", Doc: "CreateAccounts makes an account, active, for a person the directory\nvouches for and the gateway does not know. Off, only existing accounts\nsign in this way.", Nested: ""},
{Key: "roles", Type: "map[string][]string", Doc: "Roles maps a staff role (viewer, operator, billing, admin) to the\ngroups that give it, by name (CN) or whole DN. Set at every sign-in\nand sync: someone taken out of the group loses the role.", Nested: ""},
{Key: "memberships", Type: "map[string]string", Doc: "Memberships maps a group, by CN or DN, to a place in the tree —\norg:NAME, team:ORG/TEAM, project:ORG/TEAM/PROJECT — joined at sign-in\nand kept by the sync, which also takes away a place whose group the\nperson has left.", Nested: ""},
{Key: "sync_interval", Type: "time.Duration", Doc: "SyncInterval is how often the leader reads the directory. Zero is\n15m; negative turns the sync off, leaving only sign-in.", Nested: ""},
{Key: "timeout", Type: "time.Duration", Doc: "Timeout bounds each round trip to a server. Zero is 10s.", Nested: ""},
}},
{Package: "config", Name: "Landing", Title: "Landing", Doc: "", Fields: []Field{
{Key: "contact", Type: "string", Doc: "Contact is where someone without a key should go — an address, a URL, a\nchannel name. Rendered as a link when it is one, as text otherwise.", Nested: ""},
{Key: "notice", Type: "string", Doc: "Notice is an optional line of your own, for whatever the reader needs to\nknow that this page cannot guess.", Nested: ""},
+1
View File
@@ -33,6 +33,7 @@ var Sources = []string{
"internal/guard/policy.go",
"internal/guard/judge.go",
"internal/config/oidc.go",
"internal/config/ldap.go",
"internal/config/firewall.go",
"internal/config/serving.go",
"internal/config/presets.go",
+6 -1
View File
@@ -43,10 +43,15 @@ const (
// CanaryRolledBack is a model's canary taken out because it served
// worse than the model it was a candidate to replace.
CanaryRolledBack = "canary_rolled_back"
// ExternalKeyRefused is an outside provider answering 401 to one of a
// model's keys: revoked, expired or mistyped. Requests go to its other
// keys meanwhile, where it has any.
ExternalKeyRefused = "external_key_refused"
)
// Kinds are every event there is.
var Kinds = []string{ModelParked, NodeLost, BudgetSpent, UpdateAvailable, JudgeUnavailable, CanaryRolledBack}
var Kinds = []string{ModelParked, NodeLost, BudgetSpent, UpdateAvailable, JudgeUnavailable, CanaryRolledBack,
ExternalKeyRefused}
// Event is one thing that happened.
type Event struct {
+11
View File
@@ -131,7 +131,18 @@ const defaultActivationWait = 45 * time.Second
// SetNotifier tells the accountant where to say a budget is spent. Nothing
// without a store, where there are no budgets.
func (r *Router) SetNotifier(n *notify.Notifier) {
r.notifier = n
if r.acct != nil {
r.acct.SetNotifier(n)
}
}
// keyRefused says that an outside provider answered 401 to one of a model's
// keys, by its endpoint, which numbers it: never the key.
func (r *Router) keyRefused(model, endpointID string) {
r.log.Warn("an outside provider refused a key", "model", model, "endpoint", endpointID)
r.notifier.Notify(notify.ExternalKeyRefused, endpointID,
"the provider of "+model+" answered 401 to the key of "+endpointID+
": revoked, expired or mistyped; replace its file on the control plane",
map[string]any{"model": model, "endpoint": endpointID})
}
+112
View File
@@ -0,0 +1,112 @@
package router
import (
"cmp"
"context"
"errors"
"fmt"
"log/slog"
"time"
"gitea.mynodes.xyz/ilker/llmgw/internal/account"
"gitea.mynodes.xyz/ilker/llmgw/internal/config"
"gitea.mynodes.xyz/ilker/llmgw/internal/directory"
storepkg "gitea.mynodes.xyz/ilker/llmgw/internal/store"
)
// Sign-in against a directory (auth.ldap), and the accounts kept in step
// with it by the leader.
// ldapDirectory is auth.ldap as the gateway uses it.
type ldapDirectory struct {
dir *directory.LDAP
keeper *directory.Keeper
every time.Duration
label string
log *slog.Logger
}
func newLDAPDirectory(cfg config.LDAP, st *storepkg.DB, log *slog.Logger) (*ldapDirectory, error) {
dir, err := directory.NewLDAP(cfg)
if err != nil {
return nil, err
}
return &ldapDirectory{
dir: dir,
keeper: &directory.Keeper{St: st, Name: "ldap", CreateAccounts: cfg.CreateAccounts,
Mapping: directory.Mapping{Roles: cfg.Roles, Memberships: cfg.Memberships}},
every: cfg.Interval(),
label: cmp.Or(cfg.Label, "Active Directory"),
log: log,
}, nil
}
// SignIn is account.Directory: the person's name and password checked by
// the directory, then their account made or kept in step.
func (d *ldapDirectory) SignIn(ctx context.Context, login, password string) (storepkg.User, error) {
id, err := d.dir.Authenticate(ctx, login, password)
switch {
case errors.Is(err, directory.ErrBadCredentials):
return storepkg.User{}, account.ErrBadCredentials
case err != nil:
d.log.Warn("the directory could not be asked about a sign-in", "err", err)
return storepkg.User{}, fmt.Errorf("the directory could not be asked: %w", err)
}
u, err := d.keeper.SignIn(ctx, id)
switch {
case errors.Is(err, directory.ErrSuspended):
return storepkg.User{}, account.ErrSuspended
case errors.Is(err, directory.ErrNoAccount), errors.Is(err, directory.ErrNoEmail):
// Said to the log, not the form: the form says what a wrong
// password says.
d.log.Info("the directory vouched for a person who has no account here", "dn", id.DN, "err", err)
return storepkg.User{}, account.ErrBadCredentials
case errors.Is(err, directory.ErrNotActive):
return storepkg.User{}, account.ErrNotApproved
case err != nil:
return storepkg.User{}, err
}
return u, nil
}
// sync reads the directory every interval, on the leader, and brings the
// accounts it keeps in step.
func (d *ldapDirectory) sync(ctx context.Context, leader func() bool) {
if d.every <= 0 {
return
}
t := time.NewTicker(d.every)
defer t.Stop()
for {
if leader() {
d.syncOnce(ctx)
}
select {
case <-ctx.Done():
return
case <-t.C:
}
}
}
func (d *ldapDirectory) syncOnce(ctx context.Context) {
ctx, cancel := context.WithTimeout(ctx, 5*time.Minute)
defer cancel()
people, err := d.dir.People(ctx, d.keeper.Mapping)
if err != nil {
d.log.Warn("could not read the directory; accounts left as they are", "err", err)
return
}
rep, err := d.keeper.Sync(ctx, people)
if err != nil {
d.log.Error("the directory sync stopped", "err", err, "people", len(people))
return
}
for _, where := range rep.Unmapped {
d.log.Warn("auth.ldap.memberships maps a group to a place the tree does not have", "place", where)
}
if rep.Suspended+rep.Reinstated+rep.Updated > 0 {
d.log.Info("accounts kept in step with the directory", "people", len(people), "accounts", rep.Seen,
"suspended", rep.Suspended, "reinstated", rep.Reinstated, "roles_changed", rep.Updated)
}
}
+69
View File
@@ -6,11 +6,15 @@ import (
"io"
"net/http"
"net/http/httptest"
"strings"
"sync/atomic"
"testing"
"gitea.mynodes.xyz/ilker/llmgw/internal/config"
"gitea.mynodes.xyz/ilker/llmgw/internal/control/catalog"
"gitea.mynodes.xyz/ilker/llmgw/internal/notify"
"gitea.mynodes.xyz/ilker/llmgw/internal/notify/notifytest"
"gitea.mynodes.xyz/ilker/llmgw/internal/router/breaker"
"gitea.mynodes.xyz/ilker/llmgw/internal/router/endpoint"
)
@@ -87,3 +91,68 @@ func TestAnExternalModelIsNotAModelWithProfiles(t *testing.T) {
}
}
}
// Two keys for one provider, one of them at its limit, or refused: every
// request is answered by the other key, the failing one's breaker opens, and
// a refused key is said, by its endpoint and never its value.
func TestAKeyAtItsLimitOrRefusedIsPassedOverForAnother(t *testing.T) {
for _, status := range []int{http.StatusTooManyRequests, http.StatusUnauthorized} {
provider := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.Header.Get("Authorization") == "Bearer key-a" {
http.Error(w, "no", status)
return
}
w.Header().Set("Content-Type", "application/json")
fmt.Fprint(w, `{"model":"m","choices":[{"message":{"content":"ok"}}],"usage":{"prompt_tokens":1,"completion_tokens":1}}`)
}))
t.Cleanup(provider.Close)
cfg := baseConfig()
cfg.Models = []catalog.Model{{Name: "cloud", Source: "external",
External: &catalog.External{URL: provider.URL, Model: "m"}}}
rt := newTestRouter(t, cfg)
key := func(id, k string) endpoint.Info {
return endpoint.Info{ID: id, URL: provider.URL, Models: []string{"cloud"}, APIKey: k,
Engine: endpoint.EngineExternal, Pool: "external", Weight: 1, UpstreamModel: "m"}
}
rt.SetEndpoints(t.Context(), []endpoint.Info{key("external:cloud#1", "key-a"), key("external:cloud#2", "key-b")})
n, events := notifytest.Catch(t)
rt.SetNotifier(n)
// Until the failing key's breaker opens: how many requests the
// scheduler sends it before then is its own choice.
open := func() bool {
for _, s := range rt.reg.Snapshots("cloud") {
if s.ID == "external:cloud#1" && s.Breaker != breaker.Closed {
return true
}
}
return false
}
for i := 0; i < 100 && !open(); i++ {
if rec := post(t, rt.Handler(), map[string]any{"model": "cloud", "messages": []any{}}, nil); rec.Code != http.StatusOK {
t.Fatalf("key-a answering %d: request %d got %d: %s", status, i, rec.Code, rec.Body.String())
}
}
for _, s := range rt.reg.Snapshots("cloud") {
if s.ID == "external:cloud#1" && s.Breaker == breaker.Closed {
t.Errorf("%d: the failing key's breaker is still closed", status)
}
if s.ID == "external:cloud#2" && s.Breaker != breaker.Closed {
t.Errorf("%d: the good key's breaker opened", status)
}
}
var refused []notify.Event
for _, e := range events() {
if e.Kind == notify.ExternalKeyRefused {
refused = append(refused, e)
}
}
switch {
case status == http.StatusTooManyRequests && len(refused) != 0:
t.Errorf("a key at its limit was said to be refused: %+v", refused)
case status == http.StatusUnauthorized && (len(refused) != 1 || refused[0].Subject != "external:cloud#1" ||
strings.Contains(refused[0].Message, "key-a")):
t.Errorf("a refused key was said as %+v", refused)
}
}
}
+3
View File
@@ -243,6 +243,9 @@ func (r *Router) Handler() http.Handler {
return web.ModelFacts{ContextLength: m.ContextLength, MaxOutput: m.MaxOutput, Tools: m.Capabilities.Tools}
})
pages.SetPersonalOff(r.personalOff)
if r.directory != nil {
pages.SetDirectory(r.directory.label)
}
pages.SetLlamaCppVersion(r.llamaVersion)
pages.SetPersonalDefaults(r.personalDefaults)
if r.store != nil {
+25
View File
@@ -34,6 +34,7 @@ import (
"gitea.mynodes.xyz/ilker/llmgw/internal/control/catalog"
"gitea.mynodes.xyz/ilker/llmgw/internal/control/reconciler"
"gitea.mynodes.xyz/ilker/llmgw/internal/guard"
"gitea.mynodes.xyz/ilker/llmgw/internal/notify"
"gitea.mynodes.xyz/ilker/llmgw/internal/router/admission"
"gitea.mynodes.xyz/ilker/llmgw/internal/router/breaker"
"gitea.mynodes.xyz/ilker/llmgw/internal/router/endpoint"
@@ -116,6 +117,10 @@ type Router struct {
// federation is which cluster this is, to its peers; nil until said.
federation atomic.Pointer[federationID]
acct *accounting.Accountant
// notifier says what the operator must act on; nil says nothing.
notifier *notify.Notifier
// directory is auth.ldap's: sign-in and the accounts kept in step.
directory *ldapDirectory
// guard judges requests and answers; nil judges nothing. guardLibrary
// is the policies it was built from, for the pages that show them.
guard guard.Guard
@@ -392,6 +397,14 @@ func New(cfg config.Server, log *slog.Logger) (*Router, error) {
r.accounts = account.New(st, account.Config{
SignupOpen: cfg.Auth.SignupOpen,
}, log)
if cfg.Auth.LDAP.Enabled() {
d, err := newLDAPDirectory(cfg.Auth.LDAP, st, log)
if err != nil {
return nil, err
}
r.directory = d
r.accounts.SetDirectory(d)
}
r.acct = accounting.New(
accounting.NewPricing(prices, accounting.Price{
@@ -532,6 +545,9 @@ func (r *Router) Start(ctx context.Context) {
if r.threads != nil {
go r.pruneThreads(ctx)
}
if r.directory != nil {
go r.directory.sync(ctx, r.leading)
}
if r.store != nil && r.usageRetention > 0 {
go r.pruneUsage(ctx, r.usageRetention)
}
@@ -1193,6 +1209,12 @@ func (r *Router) dispatchOne(ctx context.Context, w http.ResponseWriter, req *pa
// restart. Not relayed; loaded again and tried again.
preq.RetryStatus = http.StatusNotFound
}
// One of an outside provider's several keys refused: the request
// is the provider's to answer with another of them, and the key's
// breaker counts it. With one key the 401 is the client's to see.
if dec.Endpoint.Engine == endpoint.EngineExternal && strings.Contains(dec.Endpoint.ID, "#") {
preq.RetryStatus = http.StatusUnauthorized
}
if dec.Endpoint.Peer != "" {
preq.Extra = r.peerHeaders(ctx, dec.Endpoint, req.Header)
}
@@ -1209,6 +1231,9 @@ func (r *Router) dispatchOne(ctx context.Context, w http.ResponseWriter, req *pa
} else {
out, perr = r.px.Do(uctx, w, preq)
}
if dec.Endpoint.Engine == endpoint.EngineExternal && out.Status == http.StatusUnauthorized {
r.keyRefused(model, dec.Endpoint.ID)
}
releaseAdapter()
uspan.SetAttributes(attribute.Int("http.response.status_code", out.Status))
endSpan(uspan, perr)
+2
View File
@@ -27,6 +27,8 @@ var copyOrder = []string{
"memberships",
// After users, which it points at.
"user_excluded_licences",
// After users, which it points at.
"user_directory",
"preset_state",
"api_keys",
"key_claims",
+92
View File
@@ -0,0 +1,92 @@
package store
import (
"context"
"database/sql"
"errors"
"time"
)
// DirectoryLink is an account kept by an outside directory: which one, the
// account's id there, and whether the directory's sync is what suspended it.
type DirectoryLink struct {
UserID string
Directory string // "ldap"
ExternalID string // an AD objectGUID
// SuspendedBySync is an account suspended because the directory
// disabled or lost it, and so reinstated when the directory has it
// enabled again. One an administrator suspended is not.
SuspendedBySync bool
SyncedAt time.Time
}
// DirectoryUser is the account a directory knows by that id.
func (s *DB) DirectoryUser(ctx context.Context, directory, externalID string) (DirectoryLink, error) {
var l DirectoryLink
var suspended int
var synced int64
err := s.db.QueryRowContext(ctx,
`SELECT user_id, directory, external_id, suspended_by_sync, synced_at FROM user_directory
WHERE directory = ? AND external_id = ?`, directory, externalID).
Scan(&l.UserID, &l.Directory, &l.ExternalID, &suspended, &synced)
if errors.Is(err, sql.ErrNoRows) {
return DirectoryLink{}, ErrNotFound
}
if err != nil {
return DirectoryLink{}, err
}
l.SuspendedBySync, l.SyncedAt = suspended != 0, directoryTime(synced)
return l, nil
}
// LinkDirectory records that a directory keeps an account, by its id there.
func (s *DB) LinkDirectory(ctx context.Context, userID, directory, externalID string, at time.Time) error {
_, err := s.db.ExecContext(ctx,
`INSERT INTO user_directory (user_id, directory, external_id, synced_at) VALUES (?, ?, ?, ?)
ON CONFLICT (user_id) DO UPDATE SET directory = excluded.directory, external_id = excluded.external_id,
synced_at = excluded.synced_at`,
userID, directory, externalID, at.UnixMilli())
return err
}
// DirectoryLinks are every account a directory keeps.
func (s *DB) DirectoryLinks(ctx context.Context, directory string) ([]DirectoryLink, error) {
rows, err := s.db.QueryContext(ctx,
`SELECT user_id, directory, external_id, suspended_by_sync, synced_at FROM user_directory WHERE directory = ?`,
directory)
if err != nil {
return nil, err
}
defer func() { _ = rows.Close() }()
var out []DirectoryLink
for rows.Next() {
var l DirectoryLink
var suspended int
var synced int64
if err := rows.Scan(&l.UserID, &l.Directory, &l.ExternalID, &suspended, &synced); err != nil {
return nil, err
}
l.SuspendedBySync, l.SyncedAt = suspended != 0, directoryTime(synced)
out = append(out, l)
}
return out, rows.Err()
}
// SetSyncSuspended says whether the directory's sync is what has an account
// suspended, and when it last looked.
func (s *DB) SetSyncSuspended(ctx context.Context, userID string, suspended bool, at time.Time) error {
v := 0
if suspended {
v = 1
}
_, err := s.db.ExecContext(ctx,
`UPDATE user_directory SET suspended_by_sync = ?, synced_at = ? WHERE user_id = ?`, v, at.UnixMilli(), userID)
return err
}
func directoryTime(ms int64) time.Time {
if ms == 0 {
return time.Time{}
}
return time.UnixMilli(ms)
}
@@ -0,0 +1,14 @@
-- Ported from the SQLite migration of the same number.
-- Which directory keeps an account, and its id there: an Active Directory
-- objectGUID, so a person whose mail changes there is still the same account
-- here. suspended_by_sync is an account the directory sync suspended,
-- because it was disabled or gone there; one an administrator suspended
-- here is not, and is never reinstated by the sync.
CREATE TABLE user_directory (
user_id TEXT PRIMARY KEY REFERENCES users(id) ON DELETE CASCADE,
directory TEXT NOT NULL,
external_id TEXT NOT NULL,
suspended_by_sync INTEGER NOT NULL DEFAULT 0,
synced_at BIGINT NOT NULL DEFAULT 0,
UNIQUE (directory, external_id)
);
@@ -0,0 +1,13 @@
-- Which directory keeps an account, and its id there: an Active Directory
-- objectGUID, so a person whose mail changes there is still the same account
-- here. suspended_by_sync is an account the directory sync suspended,
-- because it was disabled or gone there; one an administrator suspended
-- here is not, and is never reinstated by the sync.
CREATE TABLE user_directory (
user_id TEXT PRIMARY KEY REFERENCES users(id) ON DELETE CASCADE,
directory TEXT NOT NULL,
external_id TEXT NOT NULL,
suspended_by_sync INTEGER NOT NULL DEFAULT 0,
synced_at INTEGER NOT NULL DEFAULT 0,
UNIQUE (directory, external_id)
);
+9 -2
View File
@@ -179,10 +179,17 @@ func (s *DB) ApproveUser(ctx context.Context, id, projectID, by string, at time.
// signup suspended while pending and then reinstated would become active
// without anybody approving it.
func (s *DB) SetUserState(ctx context.Context, id string, state UserState) error {
// An account nobody approved is approved, not set active: that is what
// records who let it in. One an identity provider, SCIM or a directory
// made — active from the start, its password kept there ("!oidc",
// "!scim", "!ldap"), with no approval to record — is set active again
// when its provider reinstates it; refusing that left such accounts
// suspended for good.
if state == UserActive {
var approved sql.NullInt64
if err := s.db.QueryRowContext(ctx, `SELECT approved_at FROM users WHERE id = ?`, id).
Scan(&approved); err == nil && !approved.Valid {
var hash string
if err := s.db.QueryRowContext(ctx, `SELECT approved_at, password_hash FROM users WHERE id = ?`, id).
Scan(&approved, &hash); err == nil && !approved.Valid && !strings.HasPrefix(hash, "!") {
return errors.New("this account was never approved; approve it instead")
}
}
+24 -7
View File
@@ -86,6 +86,9 @@ type AccountPages struct {
hasQuotas bool
// oidc is sign-in through an identity provider; nil when there is none.
oidc *oidcLogin
// directory is the label of the directory that checks passwords
// (auth.ldap); empty when there is none.
directory string
// audit records every change made through these pages; nil records none.
audit audit.Sink
// hasGuard is set when /admin/guard is registered, and guardLibrary
@@ -250,6 +253,10 @@ func (p *AccountPages) signup(w http.ResponseWriter, r *http.Request) {
// --- sign in and out ---
// ssoLabel is what the sign-in button names, or "" without one.
// SetDirectory says passwords are also checked by a directory, under this
// label, so the sign-in form takes a username as well as a mail.
func (p *AccountPages) SetDirectory(label string) { p.directory = label }
func (p *AccountPages) ssoLabel(l *Lang) string {
if p.oidc == nil {
return ""
@@ -265,13 +272,22 @@ type loginData struct {
Version string
// SSO is the identity provider's label, when sign-in through one is on.
SSO string
// Directory is the directory's label when passwords are checked by one
// (auth.ldap): the field then takes a username too, not only a mail.
Directory string
// CSRF is the form's token, matched by a cookie.
CSRF string
}
// loginPage is what every rendering of the sign-in form carries.
func (p *AccountPages) loginPage(w http.ResponseWriter, r *http.Request, l *Lang, email string) loginData {
return loginData{SSO: p.ssoLabel(l), Directory: p.directory, Email: email, SignupOpen: p.accounts.SignupOpen(),
Forgot: p.resets != nil, Version: p.version, CSRF: setLoginToken(w, r)}
}
func (p *AccountPages) loginForm(w http.ResponseWriter, r *http.Request) {
noStore(w)
render(w, r, "login", loginData{SSO: p.ssoLabel(langOf(r)), SignupOpen: p.accounts.SignupOpen(), Forgot: p.resets != nil, Version: p.version, CSRF: setLoginToken(w, r)})
render(w, r, "login", p.loginPage(w, r, langOf(r), ""))
}
func (p *AccountPages) login(w http.ResponseWriter, r *http.Request) {
@@ -281,15 +297,16 @@ func (p *AccountPages) login(w http.ResponseWriter, r *http.Request) {
if !loginTokenOK(r) {
// Not this site's form: another one posting a visitor into an
// account of its choosing, or a form left open past its hour.
renderStatus(w, r, http.StatusBadRequest, "login", loginData{SSO: p.ssoLabel(l), Email: email, SignupOpen: p.accounts.SignupOpen(), Version: p.version,
Error: l.T("login.err.form_expired"), CSRF: setLoginToken(w, r)})
d := p.loginPage(w, r, l, email)
d.Error = l.T("login.err.form_expired")
renderStatus(w, r, http.StatusBadRequest, "login", d)
return
}
ctx := account.WithClient(r.Context(), clientAddr(r))
token, csrf, err := p.accounts.SignIn(ctx, email, r.FormValue("password"))
if err != nil {
d := loginData{SSO: p.ssoLabel(l), Email: email, SignupOpen: p.accounts.SignupOpen(), Forgot: p.resets != nil, Version: p.version, CSRF: setLoginToken(w, r)}
d := p.loginPage(w, r, l, email)
switch {
case errors.Is(err, account.ErrBadCredentials),
errors.Is(err, account.ErrNotApproved),
@@ -309,9 +326,9 @@ func (p *AccountPages) login(w http.ResponseWriter, r *http.Request) {
if p.oidc != nil && p.oidc.cfg.NoPasswords {
if _, u, err := p.accounts.Resolve(ctx, token); err != nil || !u.IsAdmin {
_ = p.accounts.SignOut(ctx, token)
renderStatus(w, r, http.StatusForbidden, "login", loginData{SSO: p.ssoLabel(l), Email: email,
SignupOpen: p.accounts.SignupOpen(), Version: p.version, CSRF: setLoginToken(w, r),
Error: l.T("login.err.passwords_off", p.ssoLabel(l))})
d := p.loginPage(w, r, l, email)
d.Error = l.T("login.err.passwords_off", p.ssoLabel(l))
renderStatus(w, r, http.StatusForbidden, "login", d)
return
}
}
+2
View File
@@ -19,6 +19,8 @@
"login.title": "Sign in · llmgw",
"login.heading": "Sign in",
"login.email": "Email",
"login.username": "Username or email",
"login.with_directory": "Sign in with your %s account.",
"login.password": "Password",
"login.submit": "Sign in",
"login.with_sso": "Sign in with %s",
+2
View File
@@ -19,6 +19,8 @@
"login.title": "Giriş yap · llmgw",
"login.heading": "Giriş yapın",
"login.email": "E-posta",
"login.username": "Kullanıcı adı veya e-posta",
"login.with_directory": "%s hesabınızla giriş yapın.",
"login.password": "Parola",
"login.submit": "Giriş yap",
"login.with_sso": "%s ile giriş yap",
+33 -5
View File
@@ -148,8 +148,9 @@ func (p *AccountPages) oidcStart(w http.ResponseWriter, r *http.Request) {
pr, err := o.providerFor(r.Context())
if err != nil {
p.log.Warn("the identity provider did not answer", "issuer", o.cfg.Issuer, "err", err)
renderStatus(w, r, http.StatusBadGateway, "login", loginData{SignupOpen: p.accounts.SignupOpen(), Version: p.version,
CSRF: setLoginToken(w, r), SSO: o.cfg.Label, Error: langOf(r).T("sso.err.no_answer")})
d := p.loginPage(w, r, langOf(r), "")
d.Error = langOf(r).T("sso.err.no_answer")
renderStatus(w, r, http.StatusBadGateway, "login", d)
return
}
state, nonce, verifier := randomToken(), randomToken(), oauth2.GenerateVerifier()
@@ -172,8 +173,9 @@ func (p *AccountPages) oidcCallback(w http.ResponseWriter, r *http.Request) {
if errors.As(err, &ae) {
msg = ae.in(l)
}
renderStatus(w, r, http.StatusForbidden, "login", loginData{SignupOpen: p.accounts.SignupOpen(),
Version: p.version, CSRF: setLoginToken(w, r), SSO: o.cfg.Label, Error: msg})
d := p.loginPage(w, r, l, "")
d.Error = msg
renderStatus(w, r, http.StatusForbidden, "login", d)
}
c, err := r.Cookie(oidcCookie)
if err != nil {
@@ -297,7 +299,17 @@ func (p *AccountPages) oidcUser(ctx context.Context, claims map[string]any) (sto
return store.User{}, accountError{"sso.err.state", []any{email, string(u.State)}}
}
groups := claimStrings(claims[or(o.cfg.GroupsClaim, "groups")])
claim := or(o.cfg.GroupsClaim, "groups")
if groupsOverflowed(claims, claim) {
// Read as no groups, this would take an administrator's role away
// at their next sign-in: kept as they are instead, and said.
p.log.Error("the identity provider left this person's groups out of the token, as Entra ID does past 200 groups; "+
"their role and memberships are left as they were. In the app registration's token configuration, "+
"emit only the groups assigned to the application, or provision through SCIM",
"email", email, "claim", claim)
return u, nil
}
groups := claimStrings(claims[claim])
if len(o.cfg.Roles) > 0 {
role := ""
for _, r := range []string{store.RoleAdmin, store.RoleOperator, store.RoleBilling, store.RoleViewer} {
@@ -333,6 +345,22 @@ func (p *AccountPages) oidcUser(ctx context.Context, claims map[string]any) (sto
return u, nil
}
// groupsOverflowed is a token whose groups claim was too long to send:
// Entra ID then names a source to fetch it from (_claim_names), or, in a
// token from the implicit flow, says only hasgroups.
func groupsOverflowed(claims map[string]any, claim string) bool {
if _, ok := claims[claim]; ok {
return false
}
if names, ok := claims["_claim_names"].(map[string]any); ok {
if _, ok := names[claim]; ok {
return true
}
}
has, _ := claims["hasgroups"].(bool)
return has
}
// oidcPlace finds org:NAME, team:ORG/TEAM, project:ORG/TEAM/PROJECT.
func (p *AccountPages) oidcPlace(ctx context.Context, where string) (store.Scope, string, error) {
kind, path, _ := strings.Cut(where, ":")
+9
View File
@@ -165,6 +165,15 @@ func TestSigningInThroughAProvider(t *testing.T) {
t.Error("no session cookie")
}
// Past 200 groups Entra ID sends none, only where to fetch them: the
// role stays rather than being taken away as if she had no groups.
idp.claims = map[string]any{"email": "ann@example.com", "email_verified": true,
"_claim_names": map[string]any{"groups": "src1"},
"_claim_sources": map[string]any{"src1": map[string]any{"endpoint": "https://graph.windows.net/t/users/u/getMemberObjects"}}}
if rec := signIn(t, mux, idp, false); rec.Code != http.StatusSeeOther || st.roles[u.ID] != "operator" {
t.Errorf("an overflowed groups claim: %d, role %q", rec.Code, st.roles[u.ID])
}
// Refused: a state that is not this browser's, an unverified address,
// and nobody the gateway knows when it does not make accounts.
if rec := signIn(t, mux, idp, true); rec.Code != http.StatusForbidden {
+3 -1
View File
@@ -38,7 +38,9 @@
{{if .Error}}<p class="error">{{.Error}}</p>{{end}}
<form method="post" autocomplete="on">
<input type="hidden" name="csrf" value="{{.CSRF}}">
<label>{{t "login.email"}}<input type="email" name="email" value="{{.Email}}" required autocomplete="email"></label>
{{if .Directory}}<p class="sub">{{t "login.with_directory" .Directory}}</p>
<label>{{t "login.username"}}<input type="text" name="email" value="{{.Email}}" required autocomplete="username" autocapitalize="none" spellcheck="false"></label>
{{else}}<label>{{t "login.email"}}<input type="email" name="email" value="{{.Email}}" required autocomplete="email"></label>{{end}}
<label>{{t "login.password"}}<input type="password" name="password" required autocomplete="current-password"></label>
<button type="submit">{{t "login.submit"}}</button>
</form>