For the complete documentation index, see llms.txt. This page is also available as Markdown.

POST /inference/v0/runs/{org_id}/{run_id}/resume

Resume Run

post

Resume a paused run that is waiting on a human-in-the-loop response.

How to use

Read the current pause first and submit its token with the actual human answer. A RESUMED response can be nonterminal. Remote resume is only supported for out-of-band execution; SSE is not supported for trigger-backed runs. Inspect 406/409 responses before choosing a supported continuation mode.

Authentication and permissions

Required permissions: projects.run. Resource and owner access checks also apply.

Prepare request.json

Save the example body below as request.json for the curl sample. Replace example IDs, node keys, and dynamic configuration with values for your organization; these are illustrative, not pre-existing resources.

{
  "resume_token": "TOKEN_FROM_PAUSE",
  "response": "THE_HUMAN_ANSWER"
}
Authorizations
AuthorizationstringRequired

Personal API key or the endpoint-specific legacy organization API key. Legacy inference uses the public organization key; legacy analytics uses the private organization key. This is not a password-grant login flow.

Path parameters
org_idstringRequired
run_idstring · uuidRequired

The paused run ID to resume.

Query parameters
verbosebooleanOptional

Return the full model when streaming an in-process resume over SSE.

Default: true
execution_modestring · enum · nullableOptional

Set to 'remote' to watch the continuation over a WebSocket (202 + handle) instead of blocking on this request; only valid for runs that resume out of band (Temporal/trigger) — sending it for an in-process run returns 409. Omit to resume inline (JSON, or SSE when the request sends Accept: text/event-stream).

Possible values:
Body

Resume token plus the human response to inject into the paused run.

resume_tokenstringRequired
responsestringRequired
response_typestring · enum · nullableOptional

How to interpret response, mirroring the pause's response_type.

Example: textPossible values:
Responses
200

Returns the run-result envelope (same shape as a normal run): in-memory runs resume synchronously with a terminal state + outputs; out-of-band (Temporal/trigger) runs return state=RESUMED with empty outputs, and the client polls GET /runs/{org_id}/{run_id} for completion. With Accept: text/event-stream the continuation streams back as SSE for in-process and live Temporal runs (trigger-backed runs return 406).

objectOptional
post/inference/v0/runs/{org_id}/{run_id}/resume
curl
curl --fail-with-body -sS \
  "$STACKAI_BASE_URL/inference/v0/runs/${ORG_ID}/${RUN_ID}/resume" \
  -H "Authorization: Bearer $STACKAI_API_KEY" \
  -H 'Content-Type: application/json' \
  --data-binary @request.json
{}

Last updated

Was this helpful?