List Agent Models Live
Return the provider’s live model list using the caller’s API key.
The key is forwarded directly to the provider’s /v1/models endpoint
(or Ollama’s /api/tags) and not persisted. Response shape matches
GET /agents/models per provider so the FE can swap the static fallback
for the live list without changing the dropdown contract.
422 if apiKey is missing for a non-Ollama provider. 400 with the
provider’s own error message on invalid key / connection failure.
Per-tenant rate limit: 30 req/min/bucket -> 429 with Retry-After.
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
Request body for POST /agents/models/live.
Same two shapes as ValidateLlmRequest minus model - this endpoint
lists every model the caller's credentials can see, it doesn't check a
specific one. savedConnectionId bypasses the inline-apiKey
requirement identically.
capability scopes which models the response should include:
chat— runs the chat-capable + tool-capable filters; returns each model with its per-(provider, model) tuning matrix attached, since the FE renders temperature / reasoning controls off that blob.embedding— runs the embedding-capable filter; returns each model as{value, label}(no tuning matrix — embedding endpoints accept no temperature / reasoning controls).
A second validator rejects (provider, capability) pairs that contradict
:data:PROVIDER_CAPABILITIES (e.g. provider=anthropic +
capability=embedding) before the BE wastes a round-trip on a
request the matrix already pinned as impossible.
Unified LLM provider enum.
A single AgentLlmConnection row carries one provider and a set of
capabilities (chat / embedding). PROVIDER_CAPABILITIES below pins
which capabilities each provider can serve — picked by the FE Connections
drawer and re-validated server-side on every write.
anthropic, openai, openai-responses, ollama, azure, azure-openai, bedrock, qwen, openai-compatible Which model list the FE wants. chat filters to chat + tool-capable models; embedding filters to embedding-capable models. Must be in PROVIDER_CAPABILITIES[provider].
chat, embedding Raw key - never stored
512512If set, BE resolves provider/apiKey/baseUrl from the saved llmConnection by id.
100Response
Successful Response
The response is of type Response Listagentmodelslive · object.