Realises [project-unified-models-endpoint]: cortex now surfaces every
model the operator has provisioned in the catalogue, transparently
cold-loads on the first request, and routes the request once the load
is done — without per-node configuration or client awareness of which
neuron hosts what.
cortex-core changes:
- NodeState gains `discovery: Option<DiscoveryResponse>` — populated
once per neuron on first successful poll, cached forever after
(topology is invariant for a neuron process).
- ModelProfile gains `is_feasible_on(neuron, devices)` with the
pinned_on / min_devices / min_device_vram_mb logic + 5 unit tests.
- CortexModelEntry expanded with OpenAI-compatible (`id`, `object`,
`created`, `owned_by`) plus helexa-specific extension fields
(`loaded`, `feasible_on`, `locations`).
cortex-gateway changes:
- poller.rs: `maybe_poll_discovery` fetches `GET /discovery` once per
neuron and caches on NodeState.
- handlers.rs::list_models rewritten as union of (catalogue × topology
feasibility) + (currently loaded somewhere). Catalogue-defined models
surface even when not yet loaded.
- router.rs::resolve gains priority 3 (catalogue cold-load):
1. loaded somewhere → route there
2. unloaded somewhere → route + lazy load via neuron
3. in catalogue → pick feasible neuron, POST /models/load, wait,
route. Cache the new entry locally so subsequent requests skip
the poll wait.
4. else 404
- pick_feasible_neuron prefers pinned_on neurons, falls back to any
feasible one (stable by name).
- profile_to_spec translates ModelProfile → ModelSpec, picking devices
by VRAM floor and setting tensor_parallel = min_devices for multi-
device profiles.
- "already loaded" responses from neuron are tolerated (two concurrent
requests racing the same cold-load is a benign outcome).
models.example.toml rewritten to reflect the canonical helexa fleet
(beast = 2x RTX 5090, benjy = RTX 4090, quadbrat = RTX 3060) with a
working TP example (Qwen3.6-27B pinned on beast) plus single-GPU
profiles for the smaller models.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Stage 3 of the candle-native pivot. neuron now serves
POST /v1/chat/completions backed by candle's quantized_qwen3 forward
pass on a per-model serialised generation loop, returning the standard
OpenAI ChatCompletionResponse envelope.
Pipeline per request:
- Look up the LoadedModel by request.model (404 if absent).
- Apply the Qwen3 chat template across all messages.
- Tokenize, then spawn_blocking onto tokio's blocking pool to acquire
the per-model arch lock and run prefill + greedy/temperature/top-p
sampling via LogitsProcessor.
- Stop on <|im_end|>/<|endoftext|> EOS or max_tokens (finish_reason
"stop" vs "length").
- Decode with skip_special_tokens=true, build OpenAI response with
prompt/completion/total usage counts.
Supporting changes:
- HarnessRegistry now stores Arc<dyn Harness> and caches a typed
Arc<CandleHarness> so inference routes bypass dyn-Trait dispatch.
- LoadedModel.arch becomes Arc<Mutex<ModelArch>> so the lock guard
can be moved into spawn_blocking.
- NeuronState gains an Option<Arc<CandleHarness>> field for the new
inference route.
- Typed InferenceError lets the handler map ModelNotLoaded → 404 and
other failures → 500 without string-matching anyhow messages.
- stream=true returns 501 until Stage 4 wires up SSE.
- Two leftover mistral.rs string references in proxy.rs and cortex-cli
(missed during the Stage 1 sweep) are corrected here.
Three new default-feature tests cover the no-candle 503, model-not-
loaded 404, and stream=true 501 paths. The cuda-integration test from
Stage 2 still covers real load/unload; a streaming-feature gated test
exercising actual generation will arrive with Stage 4.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Stage 1 of the candle-native pivot. Replaces the external-process
harness model (mistralrs over HTTP, llamacpp placeholder) with an
in-process Harness trait whose sole implementation is candle. The
trait keeps its shape so future engines slot in additively, but
start/stop default to no-ops and HarnessConfig drops endpoint and
systemd_unit since no harness needs external supervision.
Behaviour is unchanged on the wire: load_model returns a "not
implemented yet (Stage 2)" error and list_models is empty. The
gateway-side proxy, poller, and router are untouched.
CLAUDE.md Phase 11 (llama.cpp) and Phase 12 (mistral.rs COPR) are
marked superseded; the staged plan lives in
~/.claude/plans/create-a-more-aggressive-calm-naur.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replace NodeConfig (static vram_mb, pinned) with NeuronEndpoint.
Hardware discovery and model pinning now come from neuron API and
models.toml catalogue respectively.
- config.rs: nodes -> neurons, add models_config path
- catalogue.rs: ModelProfile with pinned_on, ModelCatalogue
- poller.rs: poll neuron GET /models (ModelInfo format)
- router.rs: resolve inference endpoint via neuron GET /models/{id}/endpoint
- evictor.rs: call neuron POST /models/unload
- node.rs: remove vram_mb, pinned fields (come from discovery/catalogue)
- All 22 gateway tests updated to mock neuron API
- Remove MistralModelsResponse, ModelLifecycleRequest (no longer needed)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Emit cortex_requests_total, cortex_request_duration_seconds,
cortex_request_errors_total, and cortex_cold_starts_total with
model and node labels on every proxied request.
Add install_test_recorder() for testing metrics without HTTP listener.
Integration test verifies counters and histograms appear after proxy.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Wire up openai_to_anthropic in the /v1/messages handler: buffer
upstream OpenAI response, parse, translate to Anthropic format
(stop_reason mapping, usage field names, content blocks).
5 integration tests covering round-trip translation, system prompt,
content blocks, and error cases.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Extract public poll_once() from poll_loop() for testability.
4 tests proving the poller correctly discovers models, updates
gateway state, marks unreachable nodes unhealthy, and prunes
stale models.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Confirms the existing proxy streams SSE chunks incrementally:
- 5-chunk test with 50ms delays verifies time spread between first
and last chunk arrival (not buffered)
- Verifies data: [DONE] terminator is forwarded
No src/ changes needed — Body::from_stream(bytes_stream()) already
handles SSE correctly.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
6 tests proving the scaffold works end-to-end:
- chat completion proxied through gateway to mock backend
- /health endpoint with healthy node
- /v1/models returns seeded model list
- 404 for unknown model
- 404 when no healthy nodes available
- 400 when request body missing model field
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add .gitea/workflows/ci.yml with fmt/clippy/test on all branches
and SRPM build + COPR publish on version tags
- Add cortex.spec for Fedora RPM packaging
- Add GPL-3.0-or-later LICENSE file
- Add cortex.example.toml with generic hostnames; gitignore cortex.toml
- Scrub infrastructure-specific hostnames from README.md, CLAUDE.md,
and doc comments
- Fix unused imports and clippy warnings to pass -D warnings
- Fix missing deps (bytes, reqwest, serde_json) exposed during build
- Run cargo fmt across workspace
- Update SPDX license identifier to GPL-3.0-or-later
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
