Skip to main content
Send chat completions through the Aria Compute ROUTER using the standard OpenAI request shape. The router matches model to a configured entrypoint, then routes the request through one of three parallel layers: semantic, agent, or afm-d. Gold entrypoints include ariacompute/semantic-auto, ariacompute/agent-auto, and ariacompute/afm-d-auto.

Endpoint

POST /v1/chat/completions Data plane endpoint. Authenticate with Authorization: Bearer <router-key>.

Request body

string
required
The entrypoint name to route through (for example ariacompute/semantic-auto, ariacompute/agent-auto, or ariacompute/afm-d-auto). Must match a configured entrypoint.
array
required
Array of message objects with role and content fields, following the OpenAI chat format.
boolean
default:"false"
If true, the response is streamed as server-sent events (SSE).
number
Sampling temperature between 0 and 2.
integer
Maximum number of tokens to generate.
array
Array of tool definitions for function calling.

Response

array
Generated completion choices. Exact shape follows the standard OpenAI chat completions format.

Example (non-streaming)

Example (streaming)