Commit Graph

5 Commits

Author SHA1 Message Date
m4 bb9bed82e1 feat(runtime)!: remove auxiliary model role, resolve all roles from run snapshot
ModelRole collapses to "primary": every role (main, tool selector, memory
agents, subagents, summarizer) resolves to the snapshot's frozen primary
model, per design 6.1/8.3 — users typically configure a single usable LLM,
so compile-time auxiliary bindings were bypassing run snapshots and
mis-attributing usage. Legacy auxiliary keys in stored snapshots, registry
JSON, and thread metadata are tolerated on read and dropped.

BREAKING CHANGE: ThreadModelSelection no longer carries an auxiliary ref;
snapshot selection_hash is computed over {primary, reasoning_effort} only;
ConfigurableModelMiddleware(role="auxiliary") is rejected.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-07-23 10:18:44 +08:00
m4 dbb6b7abde feat(model-registry): add delegation-JWT auth and config/snapshot HTTP API
- BFF service token (constant-time, plaintext or SHA-256 hash) plus
  X-Evo-Actor delegation JWT verification (ES256/RS256, iss/aud, <=60s
  lifetime, required claims, thread binding) with atomic jti anti-replay
- Config API: GET/PUT /api/model-registry, credential rotation endpoint,
  GET /api/models selector; PUT runs the section 9.2 save-time checks
  inside the registry write transaction after credential writes
- Snapshot API: create/bind/delete routes delegating to SnapshotService
  with thread/deployment binding checks and 9.5 unified error payloads
- Platform security config loader (config.yaml fields), OpenAPI export
  (scripts/export_model_registry_schema.py -> model_registry/openapi.json)
- Mount new routes in langgraph_dev/http.py; retire the legacy
  GET /api/models and POST /api/runtime-snapshots handlers
- Declare PyJWT>=2.8 (previously transitive); extend the 9.5 error code
  table with the HTTP-layer codes (400/401/403/422/500)
2026-07-21 09:28:55 +08:00
m4 b1233d42dc feat(model-registry): add ModelRegistryResolver and run snapshot service
Resolver (8.1): validates provider/model/credential/capability/limits and
the 6.5 four-mode input budget, freezes ResolvedModelConfig; resolve_for_test
relaxes only the enabled-visibility check (9.4); compute_availability is the
single 4.3 six-state judgement (stale beats configured, selectable only when
enabled).

SnapshotService (8.2, shared by the Task 5 HTTP API and Task 7 local entry):
freezes both roles' full ResolvedModelConfig with adapter spec revision,
fixed reserves, capabilities, and credential revisions; selection-hash
idempotency with pre-resolution semantics; prepared(15min)/bound(+24h)/
expired/aborted lifecycle with atomic bind; binding-checked reads that
revalidate frozen spec revisions; per-call credential resolution against the
frozen revision with no in-process secret cache (5.2); public diagnostic
view limited to the 8.2 safe subset.

Store gains additive helpers (credential pointer lookup, verification
listing, active-triplet lookup, conditional bind, due-expiry sweep) and the
taxonomy gains SNAPSHOT_NOT_FOUND (404) for missing snapshots.
2026-07-21 08:33:12 +08:00
m4 c46ae17084 feat(model-registry): add EndpointPolicy and SafeHttpTransport SSRF defenses
EndpointPolicy validates provider base URLs (section 4.3): public https
endpoints with hostname and optional port pass; loopback, private,
link-local, multicast, unspecified, and cloud-metadata addresses are
denied unless the normalized URL exactly matches a registered
development_endpoints entry (no prefix or wildcard matching). URLs with
user info, fragments, or non-http(s) schemes are rejected with the new
stable 422 code ENDPOINT_NOT_ALLOWED.

SafeHttpTransport is the single network egress for adapters: a custom
httpcore NetworkBackend resolves DNS under control on every connect
(retries included), filters denied ranges, and connects directly to the
selected IP, while TLS SNI/certificate checks and the HTTP Host header
keep the original hostname. Redirects and env proxies are disabled;
every request origin re-passes URL-layer validation before any I/O.
2026-07-20 21:25:15 +08:00
m4 c21fc0a272 feat(model-registry): add RegistryV4 schema, SQLite store, and unified error codes
New independent subpackage EvoScientist/model_registry implementing the
frozen unified model configuration design (v1.1.0, sections 4.2, 4.3,
5.1, 8.2, 9.5):

- schemas.py: single Pydantic v2 RegistryV4 schema (ModelRef, ProviderConfig,
  ModelConfig, AuthConfig) plus AdapterParameterSpec/AuthSpec/ParameterRule,
  ModelAvailability, ResolvedModelConfig (no secrets), CredentialStatus
- errors.py: all 23 stable error codes from the 9.5 table with HTTP status
  mapping and the unified {code, message, details, request_id} payload
- store.py: ModelRuntimeStore over model-runtime.sqlite3 (0700 dir, 0600
  file, WAL, foreign keys, busy_timeout) with BEGIN IMMEDIATE revision CAS,
  bootstrap->active atomic transition, immutable credential versions with
  masked status, model_verifications upsert, run_runtime_snapshots partial
  unique index, delegation_jtis, and shared-storage lock probe
- hashing.py: normalized SHA-256 configuration_hash

No existing module behavior changed. 89 new tests; full suite passes
(2922 passed, 10 skipped).
2026-07-20 20:43:28 +08:00