POST /v1/chat/completions

Part of the API reference; this page is generated from the OpenAPI document that the gateway builds from its own route tables.

Provenance

• Contract version: 1.10.0
• Contract title: acumen gateway API
• Source: acumen:docs/openapi/acumen-v1.openapi.json

Operations

POST /v1/chat/completions

OpenAI-compatible chat completions

The OpenAI chat door. Routes each request across the live seats for model (or an alias: auto, cheap, fast, coding, reason, vision, long), streaming SSE frames when stream:true. The request body is forwarded as a pass-through dialect body; tools requests skip seats probed without tool support.

• Operation id: postV1ChatCompletions
• Scope: chat
• Exposure: public
• Mutating: no
• Auth: bearerAuth, apiKeyAuth

Request

• Request body (application/json, required): ChatRequest: ChatRequest (object)

Field	Type	Required	Notes
max_tokens	integer	no
messages	array of object	yes
model	string	yes
seed	integer	no
stream	boolean	no
temperature	number	no
tools	array of object	no
top_p	number	no

Responses

• 200: Completion, or an SSE stream when the request sets stream:true
  • Streaming: text/event-stream
  • application/json: ChatCompletion: ChatCompletion (object)
• 400: schema_invalid
  • application/json: OpenAIError: OpenAIError (object)
• 401: bad_credentials
  • application/json: OpenAIError: OpenAIError (object)
• 402: quota_exhausted (cool-down applied)
  • application/json: OpenAIError: OpenAIError (object)
• 403: auth_scope
  • application/json: OpenAIError: OpenAIError (object)
• 413: payload_too_large
  • application/json: OpenAIError: OpenAIError (object)
• 429: rate_limited (RPM or concurrency)
  • application/json: OpenAIError: OpenAIError (object)
• 502: upstream_error — every attempted seat failed
  • application/json: OpenAIError: OpenAIError (object)
• 503: model_no_eligible_seat
  • application/json: OpenAIError: OpenAIError (object)