Use deepseek to parse pdf files

Discussions
1

Is it possible to use deepseek into activepieces to scrape data from pdf files? Or which other very cheap alternative or free? because I have to parse very few documents, about 2-10 montlhy and they are of 3 pages and I have to parse only one of these pages…

2

Yes, and for 2 to 10 documents a month it costs next to nothing. The PDF piece can do the trimming, so the model only reads the one page you care about:

  1. PDF piece, Extract PDF Pages: pass your file and add one range with the same start and end page (for example 2 and 2). That gives you a one page PDF.
  2. PDF piece, Extract Text on that one page file.
  3. DeepSeek piece, Ask Deepseek: put the extracted text in the prompt and ask for exactly the fields you need, for example “Return only JSON with keys invoice_number, date, total”.
  4. Map those fields into Google Sheets, a table, or wherever they go next.

Two caveats. Extract Text needs PDFs with a real text layer. If your files are scans, it returns empty text and you need a vision capable model instead. And if the page you need isn’t always the same page number, extract all pages and tell the model which section to read.

(Disclosure: I’m connected to thirds.ai.)