Extract Text from Base64 PDF and DOCX Files in Jinba Flow

Use PDF_EXTRACT_TEXT or DOCX_EXTRACT_TEXT with base64 file data and pass the extracted text to later flow steps.

Use these tools when a flow already has a PDF or DOCX file encoded as a base64 string. Choose PDF_EXTRACT_TEXT for a PDF and DOCX_EXTRACT_TEXT for a Word document. Both return the extracted text as result and require no tool configuration.

Extract text from a PDF

- id: extract_pdf
  tool: PDF_EXTRACT_TEXT
  input:
    - name: pdf_base64
      value: "<base64-encoded PDF bytes>"

Replace the placeholder with the base64 string, without a data-URL prefix. A successful run returns the text of every page in result, so later steps read it as {{steps.extract_pdf.result}}; inspect it before sending it to an AI or validation step. The tool reads the PDF’s text layer and does no OCR, so a scanned, image-only PDF returns no text. If no text appears, also check that the input contains the PDF bytes rather than a URL.

Extract text from a DOCX file

- id: extract_docx
  tool: DOCX_EXTRACT_TEXT
  input:
    - name: docx_base64
      value: "<base64-encoded DOCX bytes>"

The result is the document’s paragraph text, one paragraph per line, at steps.extract_docx.result. Text inside tables, headers and footers is not included. If extraction fails, check that the input contains a valid DOCX file (not a legacy .doc) encoded as base64. There is no EXCEL_EXTRACT_TEXT tool in the current Jinba Flow tool set; use the Excel tools for spreadsheets.