Parse a document by URL
Download the document at url, parse it, and return text, table, and image chunks in reading order. Every chunk carries its page number and bounding box. Billed per page.
Authorizations
Body
Input to a parse request.
Server-side limits (page count, file size) are not user-configurable.
Unknown fields are accepted and ignored for backward compatibility.
The ocr knob is deprecated and currently has no effect.
HTTP(S) URL of the document to parse (URL route only). Input type is detected from the file itself; the extension is a hint.
Include text chunks in the response. Set false to skip text spans; table chunks (and figures, when extract_images is true) are still returned.
Include figure (image) chunks in the response. Set false to skip figure extraction; text and table chunks are still returned.
Deprecated. Accepted for backward compatibility but currently has no effect.
auto, never, force How table chunks are structured. 'html' puts an HTML table in page_content; 'cell_grid' returns true per-cell bounding boxes. Echoed on each table chunk (falls back to 'markdown' if localization fails).
markdown, cell_grid, html 'none' (default): response unchanged. 'semantic': additionally return segments — elements grouped toward chunk_size characters at semantic/structural boundaries (headings, gaps, page breaks), each segment carrying embed-ready markdown content plus member records with page numbers and bounding boxes.
none, semantic Target segment size in characters of segment content. Segments land in a +/-25% band around this target (default 750-1250). Only meaningful when chunking is enabled; validated (200-8000) only in that case and ignored otherwise.
false (default): response unchanged. true: additionally return content — the entire parsed document as a single text string, concatenated in reading order. Not reformatted as structured markdown; it is the document's own text content end to end.
Response
Successful Response