Streaming

A flow run can return its result as a single JSON response, or stream as the run progresses — one event per node, ending with the final result. Streaming is how you build live token output or a progressive trace UI.

Turning it on

Set "stream": true in the body, or send Accept: text/event-stream:

curl -N -X POST https://api.zerowidth.ai/1.0/flows/<flowUuid>/runs \
  -H "Authorization: Bearer zw_live_…" \
  -H "Content-Type: application/json" \
  -d '{"input":{"kind":"chat","messages":[{"role":"user","content":"hi"}]},"stream":true}'

The response is text/event-stream. Each event is one node lifecycle step; a terminal event carries the same payload the non-streaming response would have returned, followed by a done sentinel.

Event types

`event`	When	Key fields
`node_start`	A node is about to run	`nodeId`, `nodeType`
`node_complete`	A node finished	`nodeId`, `outputs`, `durationMs`
`node_update`	Progress mid-node — typically token deltas from a model	`nodeId`, `data`
`node_error`	A node failed	`nodeId`, `errorMessage`
`run_complete`	Terminal success / partial	`executionId`, `status`, `outputs`, `costSummary`, `timeline`, `durationMs`
`run_error`	Terminal failure	`executionId`, `error`
`done`	Sentinel — the stream is closed	—

A streaming response always carries the full per-node timeline in its terminal event, so the verbose flag is ignored when streaming.

The full request and response shape lives on the run-a-flow reference.

2 min read