Agents

OpenAI-compatible streaming surfaces for an agent workflow: the Responses API (with native sequence_number-based resume), the Chat Completions API, and the response sub-resources (cancel, download artifacts). Each agent is addressed by its workflow id. The full SSE event taxonomy is documented in the hand-written streaming guides (/streaming/responses, /streaming/chat-completions, /streaming/artifacts); the schemas here describe only the request contracts and the transport-level response shapes.

📄️ Create an agent response (OpenAI Responses API)

OpenAI Responses API-compatible endpoint. Starts a new response for the agent identified by `id` (the workflow id). When `stream: true`, returns a `text/event-stream` of OpenAI `ResponseStreamEvent` frames with native resume via `sequence_number`; when `stream` is `false` or absent, accumulates the full run and returns a single OpenAI `Response` JSON object.

📄️ Fetch or resume an agent response

Re-opens a previously-started response. Behaviour depends on `starting_after`:

📄️ Cancel an in-progress agent response

Requests cancellation of an in-progress response. The platform sets a cancel signal that the worker polls; the underlying job transitions to a terminal state and any connected SSE consumer receives a synthesized terminal frame.

📄️ Download an agent-generated artifact

Redirects to a freshly-signed, short-lived object-storage URL for a downloadable artifact (docx / xlsx / pdf) produced by the agent during the response's run. The artifact reference is surfaced live on the SSE stream as a `response.output_item.added` item of type `artifact`.

📄️ Create an agent chat completion (OpenAI Chat Completions API)

OpenAI Chat Completions-compatible endpoint for the agent identified by `id` (the workflow id). When `stream: true`, returns a `text/event-stream` of OpenAI `chat.completion.chunk` frames terminated by `data: [DONE]`; when `stream` is false/absent, returns a single `chat.completion` JSON object.