Requests

Async submissions (images with ?async=true, all videos) return a request id (req_…). These endpoints are how you track them programmatically — the console’s Requests page shows the same records.

GET /v1/requests/{id}

curl https://<your-deployment>.inference.meshive.ai/v1/requests/req_9f8e7d6c5b4a3210 \
  -H "Authorization: Bearer $MESHIVE_API_KEY"

{
  "id": "req_9f8e7d6c5b4a3210",
  "object": "inference.request",
  "modality": "image",
  "status": "succeeded",
  "output_object_key": "serverless/requests/req_9f8e7d6c5b4a3210/result_0.png",
  "output_bucket": "meshive-r2",
  "storage_provider": null,
  "output_urls": ["https://...presigned..."],
  "error_code": null,
  "error_message": null,
  "queued_at": 1780000000,
  "started_at": 1780000004,
  "completed_at": 1780000031
}

Field	Notes
`status`	`queued → running → succeeded \| failed \| canceled` — every request terminates
`output_urls`	Presigned GET links for each produced file. Only on `succeeded`, valid ~1 hour — re-fetch for fresh links
`error_code` / `error_message`	Set on `failed` — codes like `vram_exceeded`, `model_timeout` (table)
`queued_at` / `started_at` / `completed_at`	Unix seconds

GET /v1/requests

GET /v1/requests?status=failed&limit=50

Query param	Default	Notes
`status`	—	`queued` / `running` / `succeeded` / `failed` / `canceled`
`limit`	`50`	Max `200`; newest first

Returns {"object": "list", "data": [...]} of the same objects (without output_urls — fetch a request’s detail for links).

POST /v1/requests/{id}/cancel

Cancels a request while it’s still queued and returns the updated record (status: "canceled"). Cancellation fires the request.canceled webhook event.

Status	When
`200`	Canceled
`409` `request_not_cancellable`	The request is `running` or already terminal — running work can’t be canceled
`404` / `403`	Unknown id / belongs to another workspace