Create a table and start extraction
Create a new table from uploaded files or raw text. Extraction starts automatically. Provide exactly one of: file_ids or source_text.
Usage notes
- For a working extraction, provide exactly one source: a non-empty
file_idsarray or non-emptysource_text. file_urlsremains accepted as a legacy compatibility field, but the extraction worker does not download those URLs. The request can be queued without producing a valid extraction. Download the remote file and upload it withPOST /filesfirst.- Duplicate source file names are rejected. The combined extraction size limit is 512 MB.
- The selected Product Schema is copied into the table when you create it. Later changes to the reusable Product Schema do not rewrite the table’s snapshot.
- A successful request returns
201with statusqueued_for_processing. PollGET /tables/{table_id}until the status iscompleted,error, orcancelled. - Set
webhook_urlto receive the final extraction result without polling.
Examples
Authorizations
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Body
Schema defining the column structure
Table name
IDs of previously uploaded files
1Legacy compatibility field. The handler accepts and queues these URLs, but the current extraction worker does not download them, so the extraction cannot complete correctly. Download each file and use file_ids instead.
Raw text containing product data
1Extra instructions for the AI extraction
Default context for all AI enhancements on this table
URL to notify when extraction completes or fails
Response
Created table, extraction started
Raw extraction job status returned by the API.
queued, queued_for_processing, processing, processing_by_worker, completed, error, cancelled 0 <= x <= 100files, text 
