Stream Chat Progress

Stream real-time progress for a chat job using Server-Sent Events (SSE).

Opens an SSE connection that streams events as the AI processes your request. Use this with the job_id returned from POST /v1/chat/async.

Event types:

‘intermediate’: Progress updates during processing (content, sequence, timestamp).
‘document_sync’: Initial chunk-id-tagged HTML sync, emitted once at stream start when document_html was provided.
‘proposed_change, proposed_change_batch’: A document change proposed for review (when approval_mode is ‘ask_every_time’).
‘final’: Processing complete. Contains the full result with AI response and document changes.
‘usage’: Billing data emitted after ‘final’ (monthly_used, monthly_limit, monthly_remaining).
‘ui_pointer’: Side-channel hint highlighting a UI affordance the user asked about (export/download). Emit-only; the ‘final’ content also covers the answer in prose.
‘error’: An error occurred (job failed, cancelled, or not found).

Authentication: Pass a token or api_key as a query parameter (required for EventSource which cannot set headers).

GET

chat

{session_id}

stream

Stream Chat Progress

curl --request GET \
  --url https://api.example.com/v1/chat/{session_id}/stream

{
  "detail": [
    {
      "loc": [
        "<string>"
      ],
      "msg": "<string>",
      "type": "<string>"
    }
  ]
}

Headers

authorization

string | null

Path Parameters

session_id

string

required

Query Parameters

last_sequence

integer

default:0

SSE Last-Event-ID resume: only replay intermediate responses with sequence > this. Reconnects pass the highest sequence already processed so old events are NOT re-emitted; default 0 = full history.

Required range: x >= 0

job_id

string

required

Job ID to stream progress for

token

string

Authentication token (query parameter for EventSource compatibility)

api_key

string

API key (query parameter for EventSource compatibility)

Response

Successful Response

Approve or deny AI-proposed document changes one-by-one or in batch (HITL workflow).Used with chat_async when approval_mode='ask_every_time'. For each proposed change (with chunk_id and HTML diff), respond approved=true|false plus optional feedback for the AI to revise on. Approved changes apply atomically; denied changes are discarded; the AI may revise based on feedback in the next turn. Required for regulated workflows (legal, medical, compliance) where every AI edit must be reviewed before it touches the document. For single changes: set approved=true/false and optionally provide feedback. For batch decisions: provide a 'changes' array with per-change decisions. After approval, the job resumes processing and eventually reaches status=completed.

Stream Chat Progress

curl --request GET \
  --url https://api.example.com/v1/chat/{session_id}/stream

{
  "detail": [
    {
      "loc": [
        "<string>"
      ],
      "msg": "<string>",
      "type": "<string>"
    }
  ]
}