Model Catalog Queries

<!-- generated by src/lib/agentSkills/generator.ts; manual edits will be overwritten -->

Overview

Query available AI models across all configured providers. List models, resolve model aliases, and browse the full model catalog including provider-specific variants.

Authentication

All requests require a valid Bearer token or session cookie. Obtain a token via POST /api/auth/login or configure REQUIRE_API_KEY=false for local development.

Endpoints

GET /api/v1/models

List available models

Returns all models available across configured providers. The discovery filters (capability, min_context, kind, q) are opt-in: without them the response is unchanged. They narrow the caller's own (key-filtered) catalog and run before limit/after paging.

curl http://localhost:20128/api/v1/models \
  -H "Authorization: Bearer $OMNIROUTE_TOKEN"

POST /api/v1/models/query

Query a model shortlist

Filtered shortlist over the same catalog as GET /v1/models (same auth and per-key visibility). Results are ordered by owned_by, then id, and capped by limit (default 20, maximum 100).

curl -X POST http://localhost:20128/api/v1/models/query \
  -H "Authorization: Bearer $OMNIROUTE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{}'

GET /api/v1/models/health

Per-model pause state for the caller's key

Read-only liveness view over the in-memory model lockout store. ok means no pause observed by this instance (not a health verdict); connection-level pauses live outside the store read here. Additive: leaves GET /v1/models and its cache untouched. retry_after carries the shortest remaining lockout in seconds, rounded up, only when state is cooling. One documented difference from the dispatch 429 model_cooldown rule: a model with one lockout-covered connection and one connection-level cooldown still reads cooling here, while the 429 requires every connection to carry a model cooldown.