Tabularize
Submit a tabularize job.
Mints a fresh dataset_id, reserves it in datasets_on_deck
(the aggregator atomically promotes it into datasets once the
real data_uri + columns exist), creates the jobs row with
progress.total = N, and enqueues N per-doc SQS messages.
Returns (job_id, dataset_id): poll
GET /v1/results/{job_id} for terminal status, then fetch the
produced table via GET /v1/agentic/documents/tables/{dataset_id}.
Body
POST /v1/agentic/documents/tabularize -- async tabularization job.
Extracts every uploaded document against schema and assembles a row
per document. Submit, poll status, then fetch the result via
GET /v1/agentic/documents/tables/{dataset_id}.
Fan-out concurrency is governed server-side by the Lambda fleet's reserved concurrency; there is no client-supplied concurrency knob.
Response
Successful Response
POST /v1/agentic/documents/tabularize -- async submission response.
job_id is the canonical job handle. dataset_id is the id of
the dataset being produced — used to fetch the typed result via
GET /v1/agentic/documents/tables/{dataset_id} once the job
completes.