operation: "review". With no AI, Doconda returns the same
PDF with the safe fixes applied (for example, a text layer on scanned pages) and a report with every problem left
and which page it’s on.
Steps
- Store your API key in
DOCONDA_API_KEYand install the SDK (pip install docondaornpm install @doconda/sdk). - Upload the PDF with
files.upload. - Ask for the review, read the report and download the fixed PDF.
What it checks in a PDF
- Fixed automatically: scanned pages with no text (
pdf.no_text_layer, an invisible text layer is added so you can search and copy), missing title (pdf.title_missing) and missing language (pdf.language_missing). - Reported: fonts not embedded, blank pages, mixed page sizes, sideways text, images under 150 dpi, untagged PDF (accessibility), broken internal links, empty form fields and leftover comments.
- In every format, an amount in words that doesn’t match the figure (
money.words_mismatch).
What you get back
status:ready(no problems, or all fixed),ready_with_warnings(minor warnings) orneeds_review(something is left that can’t be fixed without changing your content).- The report:
ledgerwith each fix (applied or rejected, and why) andissueswith what’s left, with its pages inlocation.pages. - It never changes your text or your values: a mistyped amount is flagged, not corrected.
Limits
- Password-protected PDFs and files that aren’t the format they claim are rejected.
- Whatever is removed for safety (actions on open, embedded files…) is listed in
sanitized.removed.
FAQ
Does it use AI?
Does it use AI?
No. The review uses no AI.
Does it work for Word, PowerPoint and Excel?
Does it work for Word, PowerPoint and Excel?
Yes, with the same call. Each format has its own list of checks in Review.
What if I only want the PDF's text?
What if I only want the PDF's text?
Use
extract: it returns the PDF as Markdown, with OCR for scanned pages.