Ilker ManapandClaude Opus 5.5 ac24e204a6 Add capacity classes and preempt lower ones to place a model ...

A model's class (critical, standard, opportunistic; unset follows its
scaling) decides what may be stopped to make room for it: a lower class
for a higher, and within a class only what another grew into under load,
for a model below its minimum. Victims are tried on a copy of the
inventory and stopped only if they make the room. Preempted replicas are
drained with a short grace and their model is held back until the room is
used. ctl model resources and class move under ctl model.

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

2026-09-30 18:02:59 +02:00

..

proxy

Serve the project's website at /about, with its fonts in the binary

2026-09-29 10:44:45 +02:00

static

Reserve CPU cores and require memory for CPU models

2026-09-30 17:43:13 +02:00

templates

Add capacity classes and preempt lower ones to place a model

2026-09-30 18:02:59 +02:00

about_test.go

Describe v0.10.0

2026-09-29 21:06:42 +02:00

about.go

Serve the project's website at /about, with its fonts in the binary

2026-09-29 10:44:45 +02:00

account_test.go

Stop shipping a placeholder that looks like a value, and say what is served

2026-09-16 18:23:19 +02:00

account.go

Connect an editor from the account page, and serve the Responses API

2026-09-29 10:04:17 +02:00

admin_mirror_test.go

Keep each held row on one line

2026-09-30 10:50:02 +02:00

admin_mirror.go

Keep each held row on one line

2026-09-30 10:50:02 +02:00

admin_page.go

Draw one menu on every page

2026-09-28 17:55:36 +02:00

admin_quotas_test.go

Let a person belong to many places, or to none

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

admin_quotas.go

Give an organization's managers a page of their own

2026-09-28 19:23:12 +02:00

admin_serve_test.go

Set how a model runs from the mirror page

2026-09-28 20:34:41 +02:00

admin_serve.go

Reserve CPU cores and require memory for CPU models

2026-09-30 17:43:13 +02:00

admin_tenancy_test.go

Let a person belong to many places, or to none

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

admin_usage_test.go

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

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

admin_usage.go

Give an organization's managers a page of their own

2026-09-28 19:23:12 +02:00

admin_users.go

Give an organization's managers a page of their own

2026-09-28 19:23:12 +02:00

chart_test.go

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

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

chart.go

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

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

charts2_test.go

Write the bar chart's latency the way its table does

2026-09-28 18:11:27 +02:00

charts2.go

Tidy the dashboard and the forms around it

2026-09-28 18:08:16 +02:00

chat_test.go

Render a finished chat reply as markdown

2026-09-28 22:44:18 +02:00

chat.go

Let a person belong to many places, or to none

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

chatrender_test.go

Render a finished chat reply as markdown

2026-09-28 22:44:18 +02:00

chatrender.go

Render a finished chat reply as markdown

2026-09-28 22:44:18 +02:00

claim_test.go

Onboard users with one-time links instead of pasted keys

2026-09-16 09:11:28 +02:00

claim.go

Onboard users with one-time links instead of pasted keys

2026-09-16 09:11:28 +02:00

codeblocks_test.go

Read the filename from a path comment inside the block

2026-09-16 21:12:00 +02:00

codeblocks.go

Read the filename from a path comment inside the block

2026-09-16 21:12:00 +02:00

connect_test.go

Connect an editor from the account page, and serve the Responses API

2026-09-29 10:04:17 +02:00

connect.go

Connect an editor from the account page, and serve the Responses API

2026-09-29 10:04:17 +02:00

contact.go

Give the landing page somewhere to send people

2026-09-16 09:22:42 +02:00

dashboard.go

Keep every admin table to one line per record

2026-09-30 11:02:18 +02:00

dashcharts.go

Label chart axes in binary sizes and small percents, and windows in words

2026-09-28 23:30:01 +02:00

download_test.go

Let a reply full of code be downloaded as a zip

2026-09-16 21:06:07 +02:00

download.go

Let a reply full of code be downloaded as a zip

2026-09-16 21:06:07 +02:00

fabric_page.go

Map every InfiniBand fabric, draw it, and place replicas by it

2026-09-29 08:46:23 +02:00

landing.go

Let people sign up, be approved, and issue their own keys

2026-09-16 09:32:35 +02:00

mirror_detail_test.go

Keep each held row on one line

2026-09-30 10:50:02 +02:00

mirror_detail.go

Open a held model to its replicas, nodes and room for more

2026-09-30 10:19:31 +02:00

mirror_explore_test.go

Split the mirror page into Held, Explore and Fetch tabs

2026-09-30 08:31:28 +02:00

mirror_explore.go

Split the mirror page into Held, Explore and Fetch tabs

2026-09-30 08:31:28 +02:00

modelcard_test.go

Show a Hugging Face repository's model card on the mirror page

2026-09-28 22:03:19 +02:00

modelcard.go

Render a finished chat reply as markdown

2026-09-28 22:44:18 +02:00

nav_test.go

Draw one menu on every page

2026-09-28 17:55:36 +02:00

nav.go

Say there are no organizations rather than 404

2026-09-28 20:23:19 +02:00

node_logs_test.go

Show a node's logs on the admin page

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

node_logs.go

Draw one menu on every page

2026-09-28 17:55:36 +02:00

node_page_test.go

Give every choice-link one pill, and stop the NVLink page breaking them

2026-09-29 09:15:44 +02:00

node_page.go

Label chart axes in binary sizes and small percents, and windows in words

2026-09-28 23:30:01 +02:00

nvlink_page.go

Find multi-node NVLink domains, place across them, and show them as racks

2026-09-29 09:07:59 +02:00

org_page_test.go

Say there are no organizations rather than 404

2026-09-28 20:23:19 +02:00

org_page.go

Say there are no organizations rather than 404

2026-09-28 20:23:19 +02:00

rows_test.go

Keep every admin table to one line per record

2026-09-30 11:02:18 +02:00

selfusage_test.go

Show a person their own use and what is left of their budgets

2026-09-28 22:51:13 +02:00

selfusage.go

Show a person their own use and what is left of their budgets

2026-09-28 22:51:13 +02:00

servable_test.go

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

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

timeseries_test.go

Label chart axes in binary sizes and small percents, and windows in words

2026-09-28 23:30:01 +02:00

timeseries.go

Label chart axes in binary sizes and small percents, and windows in words

2026-09-28 23:30:01 +02:00

web_test.go

Fit the pages to a phone, and drop an empty Fabrics heading

2026-09-29 09:24:56 +02:00

web.go

Format web.go

2026-09-30 11:02:32 +02:00