Ilker ManapandClaude Opus 5.5 b986ea180d Fixes from review: setup's memory check, MCP secrets, VS Code, bandwidth ...

- server init run again on a machine already set up is not refused its
  memory (the docs fix a certificate with --host so), and judges count
  only when one will be installed (judgeComes: not without a mirror, nor
  from a bundle without it).
- A server's sealed secrets that it no longer reads are removed when it is
  put or deleted; replacing a credential replaces every header it was
  sent, rather than adding one beside the old.
- llmgw.gatewayUrl is a machine setting and restricted in untrusted
  workspaces: a cloned repository's settings cannot send a sign-in to
  another gateway.
- The bandwidth buffer is at least eight times the CPU's last cache and at
  most a quarter of free memory, else nothing is measured until there is
  room; a registration carries the figure measured so far.

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

2026-10-29 08:23:00 +01:00

..

admintoken_test.go

Guard the admin API and pprof with a token only root and the server can read

2026-10-04 10:02:44 +02:00

admintoken.go

Guard the admin API and pprof with a token only root and the server can read

2026-10-04 10:02:44 +02:00

agent.go

The nodes' agent settings kept with the cluster's, agent.yaml down to what an agent needs to start

2026-10-08 18:06:36 +02:00

answercache_test.go

Reusing answers: a response cache per project, exact match, opt-in with a TTL

2026-10-04 23:50:53 +02:00

answercache.go

Answers to similar questions, not only the same one

2026-10-25 17:47:11 +01:00

audit.go

Audit log: every change through the admin API and the administration pages, with ctl audit

2026-10-04 16:25:56 +02:00

backup_test.go

Backups reach a replicated store: pg_dump is given the cluster's authority with verify-full

2026-10-08 21:04:46 +02:00

backup.go

Backups reach a replicated store: pg_dump is given the cluster's authority with verify-full

2026-10-08 21:04:46 +02:00

backupcopy.go

server restore: an archive checked, dry-run, and put back — configuration with modes and owners, the store in SQLite or PostgreSQL; backup_copy_to copies archives off the machine

2026-10-04 23:09:59 +02:00

bench_overhead_test.go

What llmgw adds to a request, measured: ctl bench overhead

2026-10-17 00:23:26 +02:00

bench_overhead.go

What llmgw adds to a request, measured: ctl bench overhead

2026-10-17 00:23:26 +02:00

bench_test.go

ctl bench reads a guarded engine's metrics with the engine's key, or says why it cannot

2026-10-05 00:20:25 +02:00

bench.go

What llmgw adds to a request, measured: ctl bench overhead

2026-10-17 00:23:26 +02:00

bundle.go

The cluster's settings live in the store, and every change takes effect

2026-10-05 15:25:23 +02:00

capacity.go

Keep what a benchmark found out, instead of leaving it in a terminal

2026-09-27 11:51:58 +02:00

cert.go

Add a private CA and mutual TLS for the agent stream

2026-09-16 07:28:56 +02:00

clustersettings_test.go

The cluster's settings live in the store, and every change takes effect

2026-10-05 15:25:23 +02:00

clustersettings.go

How long an on-demand model holds a card is a live setting

2026-10-14 08:58:38 +02:00

comfyuigate.go

ComfyUI as an engine of the cluster's own

2026-10-10 00:39:10 +02:00

controlplanes_test.go

The probe reads the floating address from its file each time

2026-10-05 02:56:23 +02:00

controlplanes.go

The probe reads the floating address from its file each time

2026-10-05 02:56:23 +02:00

ctl_3d.go

G4: a page that makes 3D models, and ctl 3d

2026-10-10 19:16:20 +02:00

ctl_config_test.go

The guardrails' judge has its own command, and its trouble is shown

2026-10-06 10:47:37 +02:00

ctl_config.go

The cluster's settings live in the store, and every change takes effect

2026-10-05 15:25:23 +02:00

ctl_eval_test.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

ctl_eval.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

ctl_fabrictest.go

Measure two nodes' fabric rail by rail: llmgw ctl fabric-test

2026-09-29 21:29:23 +02:00

ctl_logs.go

Show a node's logs on the admin page

2026-09-28 12:27:47 +02:00

ctl_mcp.go

The company's MCP tools under the guardrails, a rate and a breaker

2026-10-20 23:00:22 +02:00

ctl_request.go

A ComfyUI model may make meshes: from a picture, with its licence inside

2026-10-10 18:30:52 +02:00

ctl_restart.go

ctl restart --max-surge and --max-unavailable

2026-10-07 13:09:10 +02:00

ctl_status_test.go

Draw the console's overview in its panel, not over the screen

2026-10-03 23:09:40 +02:00

ctl_upgrade_test.go

An upgrade that leaves nodes behind says which

2026-10-14 09:01:09 +02:00

ctl_upgrade.go

An upgrade that leaves nodes behind says which

2026-10-14 09:01:09 +02:00

ctl.go

Company MCP servers are registered with ctl mcp, and tools granted with keys limits --tools

2026-10-20 12:50:55 +02:00

ctlweights.go

Carry a download's progress to where somebody is looking at it

2026-09-27 07:49:08 +02:00

dispatcherref_test.go

Measure two nodes' fabric rail by rail: llmgw ctl fabric-test

2026-09-29 21:29:23 +02:00

doctor.go

A node's own readiness and its multi-node readiness are kept apart

2026-10-07 13:11:16 +02:00

enginebuild.go

Choose each engine's build from the node's hardware, for every GPU generation

2026-10-01 14:35:58 +02:00

enginecheck_gpu_test.go

An adopted engine's state is written without its command; engine-check finds the installed build

2026-10-07 14:26:28 +02:00

enginecheck_test.go

SGLang read in its own metric names, started serving them, and asked for its own help

2026-10-05 00:20:25 +02:00

enginecheck.go

An adopted engine's state is written without its command; engine-check finds the installed build

2026-10-07 14:26:28 +02:00

facility_test.go

Facility-aware placement: racks, PDUs, holds and power caps

2026-10-05 01:17:49 +02:00

facility.go

Facility-aware placement: racks, PDUs, holds and power caps

2026-10-05 01:17:49 +02:00

federation.go

Federation: another llmgw cluster as a provider, its models as PEER/MODEL

2026-10-05 01:21:30 +02:00

filter_test.go

Let ctl ask a question instead of printing the whole fleet

2026-09-27 07:30:51 +02:00

filter.go

Facility-aware placement: racks, PDUs, holds and power caps

2026-10-05 01:17:49 +02:00

firewallcmd_test.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

firewallcmd.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

firewallpub_test.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

firewallpub.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

guard_judge.go

How fast the judge is, measured, and said at install

2026-10-20 22:35:19 +02:00

guard.go

Guardrails turn off and on; a judge's refusal says which hazard; ctl guard judge ask

2026-10-06 12:41:08 +02:00

join.go

Offline sites and Kubernetes: llmgw bundle create, import and trust, offline.mirror, server init --bundle; a Helm chart for the control plane

2026-10-05 00:13:39 +02:00

judgekeeper_test.go

The judge init installs is no longer replaced by the leader's own

2026-10-19 20:37:10 +02:00

judgekeeper.go

The judge init installs is no longer replaced by the leader's own

2026-10-19 20:37:10 +02:00

judges_known.go

A judge the leader serves by itself is sized as init's is

2026-10-19 16:32:04 +02:00

kb.go

Training data from a collection: llmgw ctl kb dataset

2026-10-08 20:04:37 +02:00

keys_lifecycle_test.go

Key lifecycle: lifetime limits by scope, rotation with overlap, kinds and expiry in the listing

2026-10-04 16:45:05 +02:00

keys.go

keys limits changes only the limits it is given

2026-10-20 17:32:43 +02:00

leaderforward_test.go

A standby's portal is the leader's; the changelog as the release notes read it

2026-10-05 06:34:57 +02:00

leaderforward.go

A standby's portal is the leader's; the changelog as the release notes read it

2026-10-05 06:34:57 +02:00

memory.go

Manage people and the tenancy tree, from the web and the CLI

2026-09-28 16:26:39 +02:00

mig_test.go

MIG: a card in MIG mode is placed as its instances, each a GPU of class <card>-mig-<profile>

2026-10-05 00:01:14 +02:00

mirror_gate_test.go

Release v0.10.4: CPU threads, tracing, and the rest of the review

2026-09-30 04:00:35 +02:00

mirror_import_test.go

TensorRT-LLM as an engine, and models from Ollama into the mirror

2026-10-05 00:02:01 +02:00

mirror_import.go

TensorRT-LLM as an engine, and models from Ollama into the mirror

2026-10-05 00:02:01 +02:00

mirror_test.go

Hide the mirror's bookkeeping, bound GGUF array nesting, and forget old serve jobs and downloads

2026-10-04 10:56:16 +02:00

mirror.go

TensorRT-LLM as an engine, and models from Ollama into the mirror

2026-10-05 00:02:01 +02:00

mirrorinit.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

model_canary.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

model_presets_test.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

model_presets.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

model_serve.go

A big model's download no longer runs out its startup timeout

2026-10-20 09:45:31 +02:00

model_update.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

model.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

modelplan.go

A plan for a big model: experts spread, any number of nodes, memory said

2026-10-20 08:45:33 +02:00

movestore.go

Move a store to another one without reissuing every key

2026-09-25 16:28:38 +02:00

node.go

The mirror's address is one the nodes reach, and node add checks it

2026-10-06 09:36:09 +02:00

notify.go

Notifications: webhooks for a parked model, a lost node, a spent budget and new weights; Alertmanager receiver templates

2026-10-04 16:58:11 +02:00

org_test.go

keys limits changes only the limits it is given

2026-10-20 17:32:43 +02:00

org.go

Let a person belong to many places, or to none

2026-09-28 19:03:46 +02:00

quota.go

Set capacity quotas, class, owner and reservation from the admin pages

2026-09-30 22:40:45 +02:00

release.go

Sign releases, and install only signed ones once a key is trusted

2026-09-30 23:03:48 +02:00

requestlog.go

Forgetting a person and keeping requests in the cluster: users forget, keys residency --in-cluster

2026-10-04 23:27:31 +02:00

resolve.go

Manage people and the tenancy tree, from the web and the CLI

2026-09-28 16:26:39 +02:00

restore_test.go

server restore: an archive checked, dry-run, and put back — configuration with modes and owners, the store in SQLite or PostgreSQL; backup_copy_to copies archives off the machine

2026-10-04 23:09:59 +02:00

restore.go

server restore: an archive checked, dry-run, and put back — configuration with modes and owners, the store in SQLite or PostgreSQL; backup_copy_to copies archives off the machine

2026-10-04 23:09:59 +02:00

root.go

ComfyUI as an engine of the cluster's own

2026-10-10 00:39:10 +02:00

selfupgrade.go

An upgrade that leaves nodes behind says which

2026-10-14 09:01:09 +02:00

servable_test.go

Refuse weights the nodes' llama.cpp cannot load, before fetching them

2026-09-28 20:44:36 +02:00

server_agent_test.go

Take agent_tls.insecure as consent to an agent port without TLS too

2026-10-04 10:11:40 +02:00

server_args_test.go

A broken directory setting, or a second server, no longer takes the cluster down

2026-10-19 11:10:43 +02:00

server.go

The cluster as a read-only MCP server, llmgw:admin

2026-10-20 17:18:28 +02:00

serveradd.go

A floating address keeps to its own network, and Caddy follows the leader

2026-10-15 17:35:10 +02:00

serverinit_bundle_test.go

Offline sites and Kubernetes: llmgw bundle create, import and trust, offline.mirror, server init --bundle; a Helm chart for the control plane

2026-10-05 00:13:39 +02:00

serverinit_bundle.go

The cluster's settings live in the store, and every change takes effect

2026-10-05 15:25:23 +02:00

serverinit_choices_test.go

A three-machine install can give the judges machines of their own

2026-10-19 17:45:00 +02:00

serverinit_choices.go

Fixes from review: setup's memory check, MCP secrets, VS Code, bandwidth

2026-10-29 08:23:00 +01:00

serverinit_cluster_test.go

A three-machine install can give the judges machines of their own

2026-10-19 17:45:00 +02:00

serverinit_cluster.go

A three-machine install can give the judges machines of their own

2026-10-19 17:45:00 +02:00

serverinit_firewall.go

The firewall setting reaches the mirror machine and the store's members

2026-10-09 08:53:17 +02:00

serverinit_judges_test.go

A three-machine install can give the judges machines of their own

2026-10-19 17:45:00 +02:00

serverinit_judges.go

How fast the judge is, measured, and said at install

2026-10-20 22:35:19 +02:00

serverinit_memory_test.go

Fixes from review: setup's memory check, MCP secrets, VS Code, bandwidth

2026-10-29 08:23:00 +01:00

serverinit_notify_test.go

Setup asks where alerts go, and sends each a test

2026-10-14 13:55:48 +02:00

serverinit_notify.go

Setup asks where alerts go, and sends each a test

2026-10-14 13:55:48 +02:00

serverinit_patroni.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

serverinit_postgres_test.go

Find the PostgreSQL a machine has, rather than assume its names

2026-09-28 14:47:55 +02:00

serverinit_postgres.go

The store's PostgreSQL sized to its machine's memory

2026-10-08 21:48:53 +02:00

serverinit_standby.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

serverinit_test.go

The mirror's address is one the nodes reach, and node add checks it

2026-10-06 09:36:09 +02:00

serverinit.go

Fixes from review: setup's memory check, MCP secrets, VS Code, bandwidth

2026-10-29 08:23:00 +01:00

service_test.go

llmgw start --wait waits for the models, not for the setting

2026-10-07 21:43:45 +02:00

service.go

llmgw stop, start and restart act on the whole system

2026-10-07 18:01:45 +02:00

simengine.go

Onboard a node with a script, and find two examples that never worked

2026-09-16 12:28:36 +02:00

sourcenote_test.go

Say when a replica is running an edited profile, and check a source from the web too

2026-09-27 22:42:26 +02:00

sourcenote.go

Say when a replica is running an edited profile, and check a source from the web too

2026-09-27 22:42:26 +02:00

status_test.go

An on-demand model that is up reads as ready, not as not placed

2026-10-10 01:01:55 +02:00

status.go

llmgw stop, start and restart act on the whole system

2026-10-07 18:01:45 +02:00

storeflag_test.go

Stop calling the PostgreSQL store SQLite, and stop guessing where it is

2026-09-27 12:20:40 +02:00

storeflag.go

Stop calling the PostgreSQL store SQLite, and stop guessing where it is

2026-09-27 12:20:40 +02:00

system.go

llmgw start --wait waits for the models, not for the setting

2026-10-07 21:43:45 +02:00

threads.go

Manage people and the tenancy tree, from the web and the CLI

2026-09-28 16:26:39 +02:00

toolcheck_test.go

llmgw ctl tool-check asks with the engine's key

2026-10-08 11:24:37 +02:00

toolcheck.go

llmgw ctl tool-check asks with the engine's key

2026-10-08 11:24:37 +02:00

train.go

Trained adapters published, private to their owners like the knowledge bases they come from

2026-10-08 21:30:50 +02:00

tui_actions_test.go

Canary on the console's model menu; docs

2026-10-04 20:33:59 +02:00

tui_actions.go

The guardrails' judge has its own command, and its trouble is shown

2026-10-06 10:47:37 +02:00

tui_audit.go

Usage export as CSV: by period and by place, person, key and model, from the rollups

2026-10-04 16:49:04 +02:00

tui_chart.go

Place chart points by their time, so a window's samples fill the chart's width

2026-10-04 08:55:19 +02:00

tui_choices.go

One machine runs one judge, on up to four cores

2026-10-16 19:35:42 +02:00

tui_console.go

The company's MCP servers in the console

2026-10-28 13:23:25 +01:00

tui_docs.go

A docs portal at /docs and a client developer's guide

2026-10-04 23:46:42 +02:00

tui_federation_test.go

Federation: another llmgw cluster as a provider, its models as PEER/MODEL

2026-10-05 01:21:30 +02:00

tui_federation.go

Federation: another llmgw cluster as a provider, its models as PEER/MODEL

2026-10-05 01:21:30 +02:00

tui_form.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

tui_frame.go

Show the version top right on every screen

2026-10-04 01:18:45 +02:00

tui_guard.go

Guardrails in the console

2026-10-05 03:36:17 +02:00

tui_hf.go

A plan for a big model: experts spread, any number of nodes, memory said

2026-10-20 08:45:33 +02:00

tui_home.go

Watch the cluster and edit every setting from the console

2026-10-03 22:54:09 +02:00

tui_init_test.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

tui_init.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

tui_jobs.go

Serve and update models in the background from the console, and follow them under Jobs

2026-10-04 01:27:02 +02:00

tui_kb_test.go

The console's Knowledge screen, and the manual's page on knowledge bases

2026-10-08 13:58:12 +02:00

tui_kb.go

Collections a part of the tree answers from when a request names none

2026-10-08 15:52:43 +02:00

tui_mcp_test.go

The company's MCP servers in the console

2026-10-28 13:23:25 +01:00

tui_mcp.go

Fixes from review: setup's memory check, MCP secrets, VS Code, bandwidth

2026-10-29 08:23:00 +01:00

tui_mirror.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

tui_monitor.go

Console: rotate a key, cap key lifetimes, test the webhooks; README, onboarding and console docs caught up

2026-10-04 17:28:24 +02:00

tui_node_overview_test.go

A node's console opens on what the machine is doing

2026-10-19 12:55:27 +02:00

tui_node_overview.go

A node's console opens on what the machine is doing

2026-10-19 12:55:27 +02:00

tui_node_test.go

Let server init ask for what it needs, and node add take several nodes

2026-09-29 21:54:18 +02:00

tui_node.go

Add a standby control plane with server add, and draw the screens as windows

2026-10-03 22:29:12 +02:00

tui_people.go

Answers to similar questions, not only the same one

2026-10-25 17:47:11 +01:00

tui_presets_test.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

tui_presets.go

The cluster page and the console say which picture and 3D models are here, and add them

2026-10-13 20:48:40 +02:00

tui_quality_test.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

tui_quality.go

Quality suites, and new weights held to one: ctl eval, model update/promote --eval, model gate

2026-10-05 00:04:43 +02:00

tui_setup_test.go

Setup lets every part be chosen, each with its settings; three machines each run the gateway

2026-10-15 11:22:36 +02:00

tui_setup.go

A machine too small for the models it would serve is said so first

2026-10-25 16:19:00 +01:00

tui_signin_test.go

Sign-in is set up from the console, and a directory tried before it is kept

2026-10-19 12:55:27 +02:00

tui_signin.go

Sign-in is set up from the console, and a directory tried before it is kept

2026-10-19 12:55:27 +02:00

tui_widgets.go

Open setup screens or a console when llmgw runs bare in a terminal

2026-10-03 21:33:59 +02:00

usage.go

Usage export as CSV: by period and by place, person, key and model, from the rollups

2026-10-04 16:49:04 +02:00

users.go

A forgotten password is set through a one-time link

2026-10-07 13:20:11 +02:00

wipe_test.go

The firewall turns off and on with the cluster's settings; llmgw wipe

2026-10-06 08:00:06 +02:00

wipe.go

The firewall turns off and on with the cluster's settings; llmgw wipe

2026-10-06 08:00:06 +02:00