Stop an in-flight streaming response

POST

/response-service/api/v2/agents/{agent_id}/responses/{run_id}/stop-requests

Create a stop-request for the in-flight streaming response identified by run_id. The run halts at the next graph step boundary; partial output already streamed to the client is preserved, the chat thread's checkpoint reflects state up to the last completed node, and the response log is updated with stopped_by_user=True.

Returns 202 Accepted immediately. The actual stop happens asynchronously on whichever worker is running the producer. If the producer is stuck inside a long await (e.g. mid-LLM-call), a 5s hard-kill fallback runs automatically.

Creating a stop-request for a run_id that is not currently running (already completed, never existed, or finished between request and delivery) is a safe no-op; the response is still 202.

Request

Path Params

Header Params

Responses

🟢202

application/json

Stop requested

Bodyapplication/json

🟠401

🟠422

🔴500

Request Example

Shell

JavaScript

Java

Swift

curl --location --globoff --request POST 'https:///response-service/api/v2/agents//responses//stop-requests' \
--header 'Ocp-Apim-Subscription-Key: ' \
--header 'Authorization: Bearer '

Response Example

202 - Example 1

{
    "data": {
        "run_id": "5f4927bcdc8b36d22135e"
    },
    "message": "Stop requested",
    "success": true
}

Create Bulk Agent Response Logs

Chat Thread