Ilker ManapandClaude Opus 5.5 7cb9d4fa01 Busy is not unfit: rank files by the hardware the cluster has, and let a model wait for its cards ...

Ranking what was free first made every file come out on the CPU on a cluster
whose one card was in use. The plan now ranks by hardware alone, free now
only breaking a tie; a model served onto cards that are busy is added and
waits for them, the job ending as waiting rather than failed.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

2026-10-04 01:17:32 +02:00

..

account

Close what a second security review found still open

2026-09-30 02:25:07 +02:00

accounting

Show each team's first-token and latency percentiles, errors and refusals

2026-10-03 19:50:16 +02:00

agent

Serve vision models with their projector and original weights on vLLM; group the mirror by version

2026-10-04 00:42:59 +02:00

auth

Let a person belong to many places, or to none

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

bench

Give every engine a key only the router holds

2026-09-30 00:38:29 +02:00

cli

Busy is not unfit: rank files by the hardware the cluster has, and let a model wait for its cards

2026-10-04 01:17:32 +02:00

config

Track the registry's new weights for served models, and move a model to them or back

2026-10-04 00:12:20 +02:00

control

Busy is not unfit: rank files by the hardware the cluster has, and let a model wait for its cards

2026-10-04 01:17:32 +02:00

docgen

Track the registry's new weights for served models, and move a model to them or back

2026-10-04 00:12:20 +02:00

doctor

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

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

gguf

Format kind_live_test.go

2026-10-01 07:20:35 +02:00

hf

Plan a repository's files against the fleet, and serve the one that suits it the way it fits

2026-10-04 01:03:48 +02:00

join

Let a machine join a setup by code, dialling out, with nothing logging in

2026-10-03 21:49:15 +02:00

mirror

Trace weights fetched before origins were kept, and let an origin be given by hand

2026-10-04 00:24:41 +02:00

modelplan

Busy is not unfit: rank files by the hardware the cluster has, and let a model wait for its cards

2026-10-04 01:17:32 +02:00

obs

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

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

onboard

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

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

pgha

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

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

preflight

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

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

promq

Show usage over time, per organization, team, project and person

2026-09-28 08:11:31 +02:00

release

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

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

router

Merge ha: replicated store, second control plane, join codes, setup and console screens

2026-10-03 22:59:49 +02:00

settings

Watch the cluster and edit every setting from the console

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

sim

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

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

sourcecheck

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

store

Merge ha: replicated store, second control plane, join codes, setup and console screens

2026-10-03 22:59:49 +02:00

tlsconf

Park a model that fails every time it starts, and say so everywhere

2026-09-30 01:21:57 +02:00

version

Switch the project language to English

2026-09-16 01:15:21 +02:00

web

Show each team's first-token and latency percentiles, errors and refusals

2026-10-03 19:50:16 +02:00