Run and result help

A run is one task you submitted. The API calls it an execution, so its ID is execution_id. Every run moves through a small set of statuses and ends in one final status. The result holds the model output. The receipt records how the run was handled, without the prompt or the output. Start with the status, then use the table of common problems.

What each live-run status means

A live submission returns queued when a model fits. If none fits, it already returns failed. Millwork charges the platform fee at the first event, accepted, and returns the fee and quota automatically when no model fits. A test run (the API calls it Echo, mode: "echo") finishes during submission without a model call or fee. A successful test returns completed and has a receipt but no result.

Status	What it means	What you can do next
`queued`	The run is waiting to start.	Check every two seconds for up to 90 seconds. Do not submit it again. You can still cancel, and no model call has started.
`running` or `progress`	A model call is in progress.	Check every two seconds for up to 90 seconds. Cancelling now does not stop a model call that has already started.
`completed`	Work finished. Final.	Fetch the result and the receipt.
`failed`	Work stopped with an error. Final.	Read the events, then the receipt, for the reason. If Millwork failed the run before any model attempt, it returned the fee and the quota count automatically.
`cancelled`	The run was stopped by you or by Millwork. Final.	Read the receipt. There is no result. If you cancelled it, the platform fee and the quota count are kept, and a model call that was already running finished anyway.
`expired`	The run passed its allowed time. Final.	Read the receipt. There is no result.

You can cancel a run while its status is queued, running, or progress. The four final statuses are completed, failed, cancelled, and expired; a cancel request sent then returns invalid_state. See when the platform fee is charged.

Common problems

What happened	What to do
The connection dropped after you submitted	Send the same body again with the same request key (the `Idempotency-Key` header). Millwork returns the run it already accepted. It does not start or charge a second run.
The run is still queued or running after 90 seconds	Stop polling, keep the run ID, and check again later. Do not create another run to replace it; both would be charged.
The run completed but the result returns `404`	Check the run ID, organization, and `mode`. A test run never has a result. A live result may have expired; compare any `retention_expires_at` value you saved. If you cannot explain the missing result, contact support with the run ID.
The run failed with `no_eligible_arms`	No saved model fit the request. The API calls a saved model an `arm`. Refresh the catalog and check that at least one saved model allows the kind of data you sent (`data_classes`). If no model attempt had started, Millwork returned the fee and the quota count automatically.
A provider attempt failed	Use its reason in Provider connection help.
The requested and reported model names differ	For a catalog-model run, compare both values in the result or receipt. A provider may report a dated version name. A different name alone does not mean another saved model ran.
The result has `model_provenance: null`	A directly saved model may have no provider-reported identity. The receipt records the selected saved model and usage. Do not infer the missing model or paying account; read How model routing works.
The cost limit was reached	The limit stops Millwork from starting another model attempt once recorded usage reaches it. A call already in progress can finish and take the total above the limit. The platform fee is separate.

Result and receipt

The result contains the model output. The receipt records how the run was handled and does not copy the prompt or the output. Keep the execution_id. It connects the status, events, result, receipt, and any support request.

Retry safely

Recover without starting the same work twice.

Stop a run

Cancel work you no longer need.

Start here

Fix a problem

Contact

Run and result help

What each live-run status means

Common problems

Result and receipt

Retry safely

Stop a run

​What each live-run status means

​Common problems

​Result and receipt

Retry safely

Stop a run

What each live-run status means

Common problems

Result and receipt