Async Jobs - Ascend Gateway V5

When to use async jobs

Chat sessions wait for the agent to complete before returning. Async jobs are fire-and-forget — use them when:

The task takes longer than 30 seconds (e.g. enriching 500 contacts)
You want a callback when the result is ready
You’re triggering work from another system (n8n, Zapier, your own backend)

Submitting a job

curl -X POST https://ascend-agent-worker.ascendgtm.workers.dev/jobs \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{
    "agent_type": "crm_enricher",
    "message": "Enrich all HubSpot contacts created in the last 7 days with company size, LinkedIn URL, and annual revenue. Prioritize accounts in the Series A–C range.",
    "callback_url": "https://your-app.com/webhooks/hermes-job"
  }'

Response (immediate):

{
  "job_id": "job_01J...",
  "status": "queued",
  "agent_type": "crm_enricher",
  "created_at": "2026-05-05T15:00:00Z"
}

Checking job status

curl https://ascend-agent-worker.ascendgtm.workers.dev/jobs/job_01J... \
  -H "Authorization: Bearer <your-token>"

{
  "job_id": "job_01J...",
  "status": "complete",
  "agent_type": "crm_enricher",
  "created_at": "2026-05-05T15:00:00Z",
  "completed_at": "2026-05-05T15:03:42Z",
  "response": "Enriched 47 contacts. 12 matched Perplexity data. 35 used Discover...",
  "tool_calls": [...],
  "iterations": 8
}

Webhook payload

When callback_url is set, Hermes POSTs the completed job result to your URL:

{
  "event": "job.complete",
  "job_id": "job_01J...",
  "agent_type": "crm_enricher",
  "status": "complete",
  "response": "...",
  "completed_at": "2026-05-05T15:03:42Z"
}

Failed jobs post "status": "error" with an error field. Implement idempotent webhook handlers — Hermes retries delivery up to 3 times with exponential backoff.

Job parameters

Parameter	Type	Required	Description
`agent_type`	string	Yes	Agent config to run
`message`	string	Yes	Job instruction
`callback_url`	string	No	Webhook URL for completion notification
`context`	object	No	Additional context passed to the agent
`max_iterations`	integer	No	Override config default

Job statuses

Status	Meaning
`queued`	Job accepted, not yet started
`running`	Agent loop executing
`complete`	Finished successfully
`max_iterations_reached`	Stopped at iteration limit; partial result available
`error`	Unrecoverable error; check `error_code`

Listing jobs

curl "https://ascend-agent-worker.ascendgtm.workers.dev/jobs?limit=20&status=complete" \
  -H "Authorization: Bearer <your-token>"

Results are paginated with a cursor field. Jobs are retained for 30 days.

Integration examples

n8n webhook trigger

In n8n, add an HTTP Request node that POSTs to /jobs, captures job_id, then uses a Webhook node configured at callback_url to receive the result.

Polling from your backend

If you can’t receive webhooks, poll /jobs/{job_id} every 10–30 seconds until status is not queued or running. Most jobs complete within 2 minutes; set a polling timeout of 10 minutes.

​When to use async jobs

​Submitting a job

​Checking job status

​Webhook payload

​Job parameters

​Job statuses

​Listing jobs

​Integration examples

​n8n webhook trigger

​Polling from your backend