First, smallest cut of a LAN companion bridge (split out of #855 per review):
a thin, additive, read-only layer so a LAN client can discover what a server
offers. No new LLM logic; auth is enforced by the existing AuthMiddleware.
- GET /api/companion/ping -- cheap auth-validated health check
- GET /api/companion/info -- server identity + capability flags
- GET /api/companion/models -- the CALLER's own model endpoints
/models scopes to the caller's real owner (the token's owner for bearer callers)
plus legacy null-owner shared rows, mirroring owner_filter, and never returns
api_key material. The owner rule lives in two pure helpers (token_owner,
owner_can_see) with direct tests proving a token for owner A cannot see owner B's
rows and that null-owner rows don't widen access.
Streamed deltas flagged thinking:true (reasoning-model traces) were being folded
into full_response and persisted as part of the assistant message, so saved
replies were polluted with the model's chain-of-thought. Forward those deltas to
the client (for a live thinking indicator) but exclude them from the accumulated
saved reply, in both chat and research-stream paths. Mirrors the existing rewrite
path's handling.
Mahdi Salmanzade