Create a batch
Submit uploaded files as one batch job. Returns immediately with status: "pending"; poll GET /v1/batches/{batch_id} for per-item progress. Pass an Idempotency-Key header to make retries safe: the same key within 3 days returns the same batch instead of creating a duplicate.
Authorizations
Headers
Body
The documents to process: {"type": "files", "file_ids": [...]} for previously-uploaded files, or {"type": "urls", "urls": [...]} to have us fetch each URL directly (no upload step; source bytes never written to our storage).
- FilesSource
- UrlSource
Reserved for future engine selection. Leave unset; only the default engine is accepted today.
64Include text chunks in each item's result. Set false to skip text spans in every item.
Include figure (image) chunks in each item's result. Set false to skip figure extraction in every item.
Deprecated. Accepted for backward compatibility but currently has no effect.
auto, never, force How table chunks are structured in each item's result. 'markdown' (default) keeps the existing markdown-derived cells; 'cell_grid' returns true per-cell bounding boxes on table chunks.
markdown, cell_grid, html 'none' (default): item results unchanged. 'semantic': each item's result additionally carries segments — elements grouped toward chunk_size characters at semantic/structural boundaries.
none, semantic Target segment size in characters. Segments land in a +/-25% band around this target. Only meaningful when chunking is enabled; validated (200-8000) only in that case and ignored otherwise.
Arbitrary JSON stored with the batch and echoed back on every poll, in the batch list, and in webhook event bodies. Put your own run ids, tenant ids, or job references here. It never affects processing. Maximum 16 KB serialized.
Optional completion webhook for this batch. Omitted means no webhook fires. {"mode": "svix"} delivers a signed batch.update event to your registered endpoints when the batch reaches a terminal status; {"mode": "direct", "url": ...} sends an unsigned event to the given HTTPS URL (prototyping).
Response
Successful Response