9ff38dfe4e
v2.3 review: drop a duplicate logIdentityMode call; README wording; review notes
v2.3
kyleandClaude Opus 5.52026-09-25 20:17:03 -07:00
7a12ddcf5a
Context refusal in llama-server's exceed_context_size_error shape; README for v2.3
kyle2026-09-25 20:15:49 -07:00
fa4d06170f
Merge remote-tracking branch 'origin/master' into v2.3
kyle2026-09-25 20:11:27 -07:00
ae7b10b4db
v2.3 task 04: replacement ctxguard_router_test.go for the new refusal body; run notes
kyleandClaude Opus 5.52026-09-25 20:11:12 -07:00
15f6c62381
v2.3/04-ctx-error-docs: stopped, protected v2.1 router test conflicts with new body
kyle2026-09-25 20:10:09 -07:00
fa1c398fc4
A route may have its own listener: every request there is that route, paths unprefixed Implemented-By: OpenCode session (model recorded in docs/implementer-log.md)
kyle2026-09-25 19:59:21 -07:00
35b07ece81
Merge remote-tracking branch 'origin/master' into v2.3
kyle2026-09-25 19:34:20 -07:00
d3f1d8e55a
AGENTS.md: never tune production timing to a test; commit messages via .state/commit-msg.txt
kyleandClaude Opus 5.52026-09-25 19:34:13 -07:00
72bc9c1a27
Merge remote-tracking branch 'origin/master' into v2.3
kyle2026-09-25 19:33:49 -07:00
aea2eeae2c
Routes may share one lease (affinity = "route") and skip crossbar's queue (queue = false)
kyleandClaude Opus 5.52026-09-25 19:33:46 -07:00
fe6cd447c8
v2.3: fix the queue=false release race in the given test; add TestLoadIsHeldForTheWholeStream; run notes
kyleandClaude Opus 5.52026-09-25 19:33:27 -07:00
33fa61bedb
Control-plane requests follow the lease but take no slot and write no row
kyle2026-09-25 17:44:00 -07:00
3518e84dd7
v2.3 plan: clients that manage their own slots (control calls, route affinity, queue = false, route listeners, llama-server ctx error); given tests
kyleandClaude Opus 5.52026-09-25 17:36:59 -07:00
4c6415819e
deploy/hyperborea: opencode-*/hermes-*/probe-* templates; titan wake to both segments
kyleandClaude Fable 5.12026-09-25 13:48:32 -07:00
11c8e053f7
Merge v2.1: cancellation recorded from the reverse proxy; router mode loaded/per-model context
kyleandClaude Fable 5.12026-09-25 11:49:45 -07:00
055ab079d4
Learn per-model context from /props?model=; only status "loaded" is loaded
kyle2026-09-25 11:47:35 -07:00
cb5678abc0
v2.1 README: note the moved-header separator gap (v2 task text ambiguous, untested)
kyleandClaude Fable 5.12026-09-25 11:45:52 -07:00
f3dfdbfa50
Record cancellation from what the reverse proxy observed, not the request context
kyle2026-09-25 11:32:18 -07:00
317293cfb3
v2.1 task 01: state the ReverseProxy facts (ErrorHandler before headers, ErrAbortHandler mid-body); note the refusal-ending
kyleandClaude Fable 5.12026-09-25 11:26:01 -07:00
d596ec8abf
v2.1 task 02: router mode — status loaded only, per-model /props, guard and hosts view use it; given tests
kyleandClaude Fable 5.12026-09-25 11:22:37 -07:00
98faa3a57d
Given tests: order arrivals by limiter state instead of sleeps (limiter, spread, cancel-while-queued)
kyleandClaude Fable 5.12026-09-25 11:20:26 -07:00
a49a86c0e4
v2.1 plan: task 01 cancel-record with its given served_test.go; index for 02/03
kyleandClaude Fable 5.12026-09-25 11:12:58 -07:00
d0d3203f73
Wire wake and identity into crossbar; v2 smoke and README
kyle2026-09-25 11:12:24 -07:00
47f49072cc
Merge remote-tracking branch 'origin/master' into v2
kyle2026-09-25 11:08:24 -07:00
481ea12f4c
v1 proxy test: order arrivals by limiter state; note the 499-on-served-200 defect for v2.1
kyleandClaude Fable 5.12026-09-25 11:08:16 -07:00
27debb9790
v1 limiter test: release before reporting in TestParallelAndQueue (race with the final count check)
kyleandClaude Fable 5.12026-09-25 11:02:07 -07:00
087323ad62
Merge remote-tracking branch 'origin/master' into v2
kyle2026-09-25 10:50:23 -07:00
7710352462
v2 smoke: send the 300 KB prompt via -d @file (argv element limit); note in README
kyleandClaude Fable 5.12026-09-25 10:49:56 -07:00
962ea2aea4
v2 plan: note the router /props autoload fact and the v2.1 follow-up
kyleandClaude Fable 5.12026-09-25 10:12:29 -07:00
bddf6c93f2
Add the wake package: magic packets and a waiter
kyle2026-09-25 10:03:07 -07:00
2fe3b9c865
Proxy: move or refuse prompts that do not fit the leased host's context
kyle2026-09-25 09:56:46 -07:00
959aec92f1
Merge origin/master into v2: sticky-growth test fixed
kyle2026-09-25 09:55:15 -07:00
6c8cbbc78d
v2 plan: sticky-growth test appends turns instead of changing the first user message; note the finding
kyleandClaude Fable 5.12026-09-25 09:54:36 -07:00
056a527fd7
Health: learn n_ctx and total_slots from /props
kyle2026-09-25 09:14:08 -07:00
9f5b50acf6
Review fixes: record cancelled requests as 499; empty usage is an array
kyle2026-09-25 09:00:31 -07:00
fbb952bae9
v1.1 plan: note the third stop (timing flakes under load, refusal-ending)
kyleandClaude Fable 5.12026-09-25 08:59:56 -07:00
d8f6d2a804
v1.1 plan: state the ReverseProxy abort behaviour the fix depends on; note the second stop
kyleandClaude Fable 5.12026-09-25 08:41:30 -07:00
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
kyleandClaude Fable 5.12026-09-25 08:37:38 -07:00
9ba04b16e0
v1.1 plan: cancelled requests recorded as 499, empty usage is []; tests proven failing on master
kyleandClaude Fable 5.12026-09-25 08:28:07 -07:00
c2136b3515
v1 plan: task 06 must update the one admin.Handler call in main.go for the gate; note the finding
kyleandClaude Fable 5.12026-09-25 06:29:09 -07:00
c2a8b88a3f
Merge origin/master into v1: task 06 split into admin + main
kyle2026-09-25 06:08:28 -07:00
4da47b3dfa
v1 plan: split task 06 into admin (06) and main wiring (07); smoke becomes 08; note the finding
kyleandClaude Fable 5.12026-09-25 06:07:53 -07:00
97f7cdffd9
Route by lease, queue per host and model, record every request
kyle2026-09-25 05:15:21 -07:00
02008f6d56
Merge origin/master into v1: given proxy tests split under the line limit
kyle2026-09-25 05:08:12 -07:00
ebcf6e4536
v1 plan: split the given proxy tests under the 400-line limit (scaffolding into helpers_test.go); note the finding
kyleandClaude Fable 5.12026-09-25 05:08:03 -07:00
740583f812
Merge origin/master into v1: proxy test fix, helpers_test.go given
kyle2026-09-25 04:56:33 -07:00
7bc6761120
v1 plan: helpers_test.go as a given file, fake upstream records /v1/models, no /tmp rule; note task 05 findings
kyleandClaude Fable 5.12026-09-25 04:56:22 -07:00