Agent Result Check
Poll this endpoint to check the status of an Agent request and retrieve the results.
Authorizations
Path Parameters
Response
Successful Response
Response model for the /agent endpoint.
The status of the request. Should be 'complete' when the request is done.
Signed URL for downloading the completed result JSON when direct result download is required.
Number of seconds until result_url expires.
Whether the agent run succeeded.
If the run was not successful, this will contain an error message.
The output shape requested for this run: 'parse', 'json', 'wcag', 'jats', or 'segment'.
Whether the agent finished its task before hitting max_turns.
The number of turns the agent took.
The markdown output of the run, when applicable.
The HTML output of the run, when applicable.
The structured JSON output of the run, when applicable.
The output in chunks format, when applicable.
The filled-in extraction schema, when applicable.
Segment records for the 'segment' output shape (production /segment sense): {name, pages (0-indexed), confidence}.
A dictionary of metadata about the document and the run.
The agent's final verdict for generic output shapes, when applicable.
The JATS 1.2 Journal-Publishing article XML, for the 'jats' output shape.
The tagged, accessible PDF (PDF/UA-1) as a base64 string, for the 'wcag' output shape.
Images produced by the run as base64 strings, keyed by the name the output references them by. Only populated when return_images was set on the request.
Verification results emitted as metrics (not a gate).
The number of pages that were processed.
The total cost of the run.
A dictionary of the cost breakdown of this request.
The parse quality score of the output, if available.
The runtime of the run.
A dictionary of the versions of the libraries used in the request.