kyle/crossbar
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
100 Commits 6 Branches 0 Tags
35b07ece81f936b8fea109390d5876af5d4d43de
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
kyle 35b07ece81 Merge remote-tracking branch 'origin/master' into v2.3 2026-09-25 19:34:20 -07:00
kyleandClaude Opus 5.5 d3f1d8e55a AGENTS.md: never tune production timing to a test; commit messages via .state/commit-msg.txt 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-25 19:34:13 -07:00
kyle 72bc9c1a27 Merge remote-tracking branch 'origin/master' into v2.3 2026-09-25 19:33:49 -07:00
kyleandClaude Opus 5.5 aea2eeae2c Routes may share one lease (affinity = "route") and skip crossbar's queue (queue = false) 
route.go gains Route.Affinity/Queue with PerRoute(), Queues() and affinity
validation (checkRoutes moved here; config.go calls it once). limiter.Track
counts a request without holding or refusing it; a release hands the slot to a
waiter only while in flight <= parallel. The proxy leases a PerRoute() route
under an empty fingerprint (the row keeps the real one) and uses Track when
Queues() is false.

Implemented by Ornith (OpenCode); owner review removed a release-on-first-flush
workaround for a race in the owner's given test (see implementer log).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-25 19:33:46 -07:00
kyleandClaude Opus 5.5 fe6cd447c8 v2.3: fix the queue=false release race in the given test; add TestLoadIsHeldForTheWholeStream; run notes 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-25 19:33:27 -07:00
kyle 33fa61bedb Control-plane requests follow the lease but take no slot and write no row 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 17:44:00 -07:00
kyleandClaude Opus 5.5 3518e84dd7 v2.3 plan: clients that manage their own slots (control calls, route affinity, queue = false, route listeners, llama-server ctx error); given tests 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-09-25 17:36:59 -07:00
kyleandClaude Fable 5.1 4c6415819e deploy/hyperborea: opencode-*/hermes-*/probe-* templates; titan wake to both segments 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 13:48:32 -07:00
kyleandClaude Fable 5.1 9872084165 Merge v2.2: route templates, multiple wake broadcast addresses 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 13:48:23 -07:00
kyleandClaude Fable 5.1 bd0a9f2ff0 v2.2: adopt example.toml as a given copy; run notes 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 13:48:16 -07:00
kyle 8018f67031 Wake: a target may list several broadcast addresses 
Config keeps broadcast (one) and adds broadcasts (a list); exactly one must
be present. Wake.Addresses() returns Broadcast then Broadcasts; checkWake
errors on both-set, neither-or-empty, and non-host:port entries. Waker sends
to every address in order, logging/past a failure so one dead address does
not stop the others, and returns false only when none could be sent.

Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 13:45:39 -07:00
kyle 4e1dd03d07 Route templates: a route named x-* serves any request route x-<something> 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 13:38:24 -07:00
kyleandClaude Fable 5.1 4863e53eb8 v2.2 task 01: put Route() in a new config/route.go (config.go is at the line limit); note 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 13:16:19 -07:00
kyleandClaude Fable 5.1 5de096a221 v2.2 plan: route templates (opencode-*) and multiple wake broadcast addresses; given tests 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 13:04:15 -07:00
kyleandClaude Fable 5.1 3ec53d52e5 deploy/hyperborea: enable titan wake block with the fixed Wi-Fi private MAC (best effort) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 12:58:32 -07:00
kyleandClaude Fable 5.1 6e9a70587a deploy/hyperborea: production config, user unit, install script, README (wake block pending titan's MAC) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 12:57:25 -07:00
kyleandClaude Fable 5.1 4f03cb2c52 README: context guard section (X-Crossbar-Ctx, 400 body), per-model context in the hosts view, router-mode note 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:50:38 -07:00
kyleandClaude Fable 5.1 11c8e053f7 Merge v2.1: cancellation recorded from the reverse proxy; router mode loaded/per-model context 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:49:45 -07:00
kyle 055ab079d4 Learn per-model context from /props?model=; only status "loaded" is loaded 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 11:47:35 -07:00
kyleandClaude Fable 5.1 cb5678abc0 v2.1 README: note the moved-header separator gap (v2 task text ambiguous, untested) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:45:52 -07:00
kyle f3dfdbfa50 Record cancellation from what the reverse proxy observed, not the request context 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 11:32:18 -07:00
kyleandClaude Fable 5.1 317293cfb3 v2.1 task 01: state the ReverseProxy facts (ErrorHandler before headers, ErrAbortHandler mid-body); note the refusal-ending 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:26:01 -07:00
kyleandClaude Fable 5.1 d596ec8abf v2.1 task 02: router mode — status loaded only, per-model /props, guard and hosts view use it; given tests 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:22:37 -07:00
kyleandClaude Fable 5.1 98faa3a57d Given tests: order arrivals by limiter state instead of sleeps (limiter, spread, cancel-while-queued) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:20:26 -07:00
kyleandClaude Fable 5.1 6c3a2cff8a Merge v2: learned context, context guard, wake-on-LAN, identity 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:14:27 -07:00
kyleandClaude Fable 5.1 a49a86c0e4 v2.1 plan: task 01 cancel-record with its given served_test.go; index for 02/03 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:12:58 -07:00
kyle d0d3203f73 Wire wake and identity into crossbar; v2 smoke and README 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 11:12:24 -07:00
kyle 47f49072cc Merge remote-tracking branch 'origin/master' into v2 2026-09-25 11:08:24 -07:00
kyleandClaude Fable 5.1 481ea12f4c v1 proxy test: order arrivals by limiter state; note the 499-on-served-200 defect for v2.1 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:08:16 -07:00
kyleandClaude Fable 5.1 27debb9790 v1 limiter test: release before reporting in TestParallelAndQueue (race with the final count check) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 11:02:07 -07:00
kyle 087323ad62 Merge remote-tracking branch 'origin/master' into v2 2026-09-25 10:50:23 -07:00
kyleandClaude Fable 5.1 7710352462 v2 smoke: send the 300 KB prompt via -d @file (argv element limit); note in README 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 10:49:56 -07:00
kyleandClaude Fable 5.1 6261b68a4b v2 plan: note the task 05 12-second stop (typo'd path, refusal-ending) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 10:18:28 -07:00
kyle a942e336d8 Add identity: whois resolver, checker, header mode, middleware; config for wake, peers, identity 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 10:16:51 -07:00
kyleandClaude Fable 5.1 962ea2aea4 v2 plan: note the router /props autoload fact and the v2.1 follow-up 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 10:12:29 -07:00
kyle bddf6c93f2 Add the wake package: magic packets and a waiter 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 10:03:07 -07:00
kyle 2fe3b9c865 Proxy: move or refuse prompts that do not fit the leased host's context 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 09:56:46 -07:00
kyle 959aec92f1 Merge origin/master into v2: sticky-growth test fixed 2026-09-25 09:55:15 -07:00
kyleandClaude Fable 5.1 6c8cbbc78d v2 plan: sticky-growth test appends turns instead of changing the first user message; note the finding 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 09:54:36 -07:00
kyle 056a527fd7 Health: learn n_ctx and total_slots from /props 
Implemented /props learning in internal/health/health.go (added Status.NCtx/Status.Slots, PerSlotCtx, and a best-effort GET <base>/props appended to the poll after /v1/models; 0/unknown on any failure without failing the poll) and exposed them in internal/admin/admin.go HostView. Copied internal/health/props_test.go and the replacement internal/proxy/helpers_test.go byte-identical to docs/plans/v2/_files/.

Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 09:14:08 -07:00
kyle f73ffb0d27 Merge origin/master into v2: helpers_test.go answers /props 2026-09-25 09:12:28 -07:00
kyleandClaude Fable 5.1 927b2cfc4d v2 plan: fake upstream answers /props without counting it (task 01 replacement helper); note the finding 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 09:11:54 -07:00
kyle 5afacbc050 Health: learn n_ctx and total_slots from /props 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 09:10:38 -07:00
kyle 55679c5678 Merge v1.1: cancelled requests recorded as 499; empty usage is [] 2026-09-25 09:02:25 -07:00
kyleandClaude Fable 5.1 6bdcf437f4 v1.1 review: checklist, probes, three findings 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 09:02:13 -07:00
kyle 9f5b50acf6 Review fixes: record cancelled requests as 499; empty usage is an array 
Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
 2026-09-25 09:00:31 -07:00
kyleandClaude Fable 5.1 fbb952bae9 v1.1 plan: note the third stop (timing flakes under load, refusal-ending) 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 08:59:56 -07:00
kyleandClaude Fable 5.1 d8f6d2a804 v1.1 plan: state the ReverseProxy abort behaviour the fix depends on; note the second stop 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 08:41:30 -07:00
kyleandClaude Fable 5.1 74b8af4ce0 AGENTS.md: a refused tool call is not a reason to end the turn; v2 plan (props, ctx guard, wake, identity, wiring) as acceptance tests; v1.1 run note 
v2 given tests compiled against a panic-only skeleton (go vet clean); no reference
implementation.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 08:37:38 -07:00
kyleandClaude Fable 5.1 9ba04b16e0 v1.1 plan: cancelled requests recorded as 499, empty usage is []; tests proven failing on master 
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
 2026-09-25 08:28:07 -07:00