Claude
6afffbcbf2
The engine already tracks where each turn's wall time goes (expert-disk service, I/O wait, expert matmul, attention, lm_head) — it just only spoke at exit or under PROF=1. Stream it instead: - glm.c: mux serve emits a per-turn "PROF" protocol line next to TIERS/HITS (window deltas per request, same convention as the STAT hit%); the phase window base is now always captured (a few loads per request). - openai_server.py: parses PROF into a 120-turn rolling window and serves it at /profile (read-only, same trust level as /health). - web: new Profiling tab — stat tiles (tok/s, wall, tokens/forward, disk service), wall-time composition bars for the last turn and the window, per-turn throughput and stacked phase columns with hover readouts, and a table of recent turns. Disk service is shown apart from the stack: it overlaps with compute, so only the I/O wait the compute thread felt counts inside wall time. Phase colours are a CVD-validated set with gaps + legend + table so identity never rides on colour alone. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WhTmF8yvEBgSkUKSVfZF7P
colibrì web
React/Vite interface for an OpenAI-compatible colibrì server.
npm install
npm run dev
The default endpoint is http://127.0.0.1:8000/v1. Start the API server from
PR #21 (or any compatible backend), then use Probe server to load its models.
Local validation:
npm test
npm run build
Besides Chat and Brain, the Profiling tab charts where the engine spent
each turn's wall time (I/O wait, expert matmul, attention, LM head) from the
server's /profile endpoint — a rolling window of per-turn PROF snapshots
emitted by the engine.
The test suite stays browser-light: API requests use a mocked fetch, while
runtime capability and storage behavior are covered through pure helpers. It
checks that /health and /profile are resolved next to (not below) the OpenAI /v1 prefix,
supports both boolean and numeric scheduler.active responses, and sends the
colibrì-specific cache_slot field only when KV-slot support was advertised.
The endpoint and selected model are persisted locally. API keys are intentionally
memory-only; startup/persistence also removes the legacy colibri.apiKey value.