Guides
Guides
These are the notes I wish a PDF had shipped with. TableJSON reads a text layer and returns rows. The playground is enough for one file. The pages below are for the second file, when the header is wrong, the statement has a blank credit column, or you are about to loop the API over a folder.
Each guide uses the real response shape: headers, raw_cells, page_number, page_end, and, for invoices, the PASS or FAIL checksum. Nothing here describes an OCR mode. There is not one. If you only remember a single check, highlight a number in the PDF before you upload it.
Extracting tables from a PDF into JSONWhat the playground and POST /v1/extract-tables return, how a line becomes cells, and how to spot a fallback line_number table.
Turn an invoice PDF into Excel or CSVThe checksum workbook, which headers map, why FAIL still downloads, and how to save the sheet as CSV without losing the vendor.
Text-layer PDFs and scanned PDFsThe highlight test, what a scan returns, and why OCR has to happen before this site ever sees the file.
Cleaning messy table JSONTitles that became headers, columns that slid left, page furniture, and a five-line check before you import.
Bank statement PDFs into a spreadsheetNo special bank parser. How blank debit and credit cells shift the balance, and the opening-balance row that double-counts.
JSON or CSV for table dataWhy the API returns JSON, when a quoted CSV is what a colleague actually wanted, and how SKUs lose their zeros.
Automating PDF table extraction with the APIKeys, the 500 MB and 50 MB limits, a folder loop, and the status fields that should stop a scan from posting.
The short version of the same facts is on the playground page, under the tool. Request fields and error codes are on the API reference.