Stream a chat response

The response is text/event-stream. Each frame contains an event field (event type) and a data field (JSON payload). Consume text_delta events to progressively render the assistant’s reply.

Event types in order:

• stream_start — stream opened, includes model and provider info
• step_start — reasoning step started
• text_start — text block started
• text_delta — incremental text chunk; read data.delta for the token
• text_complete — text block finished
• step_finish — reasoning step finished
• stream_end — stream closed, includes finish_reason and token usage

Example stream:

event: stream_start
data: {"id":"evt_01KPTDR1BWRTJRY7W56MBCGF0R","timestamp":1776855811,"model":"grok-4-1-fast-non-reasoning","provider":"xai","metadata":null}

event: step_start
data: {"id":"evt_01KPTDR1BWRTJRY7W56MBCGF0S","timestamp":1776855811}

event: text_start
data: {"id":"evt_01KPTDR1BWRTJRY7W56MBCGF0T","timestamp":1776855811,"message_id":"evt_abc"}

event: text_delta
data: {"id":"evt_01KPTDR1BXRXHBK6DFSZ3W32H6","timestamp":1776855811,"delta":"Hello","message_id":"evt_def"}

event: text_delta
data: {"id":"evt_01KPTDR1CXK80DJ647VEF68ATW","timestamp":1776855811,"delta":"!","message_id":"evt_klm"}

event: text_complete
data: {"id":"evt_01KPTDR1Q0Z330AKTGKFXYRQ7G","timestamp":1776855811,"message_id":"evt_xyz"}

event: step_finish
data: {"id":"evt_01KPTDR1Q1PD7FG8F9DS2GCAVJ","timestamp":1776855811}

event: stream_end
data: {"id":"evt_01KPTDR1Q1PD7FG8F9DS2GCAVK","timestamp":1776855811,"finish_reason":"Stop","usage":{"prompt_tokens":12,"completion_tokens":8,"cache_write_input_tokens":null,"cache_read_input_tokens":null,"thought_tokens":null},"citations":null}

Note: Use curl -N (no-buffer) or an EventSource client to see tokens arrive in real time. Postman buffers SSE until the connection closes.