SkillHub

extracting-mistral-ocr

v1.0.0

Extracts text, tables, and images from PDFs (including scanned PDFs) using the Mistral OCR API. Use when user asks to OCR a PDF/image, extract text from a PDF, parse a scanned document, convert a PDF to Markdown, or extract structured fields from a document.

Sourced from ClawHub, Authored by tristanmanchester

Installation

Please help me install the skill `extracting-mistral-ocr` from SkillHub official store. npx skills add tristanmanchester/extracting-mistral-ocr

Mistral OCR PDF extraction

Quick start (default)

Run the bundled script to OCR a local PDF and write Markdown + JSON outputs:

python {baseDir}/scripts/mistral_ocr_extract.py --input path/to/file.pdf --out out/ocr

Output directory layout:

  • combined.md (all pages concatenated)
  • pages/page-000.md (per-page markdown)
  • raw_response.json (full OCR response)
  • images/ (decoded embedded images, if requested)
  • tables/ (separate tables, if requested)

Workflow

  1. Pick input mode
  2. Local PDF (most common): upload via Files API, then OCR via file_id.

  3. Public URL: OCR directly via document_url.

  4. Choose output fidelity (defaults are safe for RAG)

  5. Keep table_format=inline unless the user explicitly wants tables split out.
  6. Set --include-image-base64 when the user needs figures/diagrams extracted.

  7. Use --extract-header/--extract-footer if header/footer noise hurts downstream search.

  8. Run OCR

  9. Use scripts/mistral_ocr_extract.py to produce a deterministic on-disk artefact set.

  10. (Optional) Structured extraction from the whole document

  11. If the user wants fields (invoice totals, contract parties, etc.), provide an annotation prompt.
  12. The OCR API can return a document-level document_annotation in addition to page markdown.

Example:

bash python {baseDir}/scripts/mistral_ocr_extract.py --input invoice.pdf --out out/invoice --annotation-prompt "Extract supplier_name, invoice_number, invoice_date (ISO-8601), currency, total_amount. Return JSON." --annotation-format json_object

Decision rules

  • If the PDF is local and not publicly accessible, upload it (the script does this automatically).
  • If the PDF URL is private or requires authentication, do not pass it as document_url; upload instead.
  • If output quality is critical, prefer table_format=html for downstream parsing over brittle regex.

Common failure modes

  • Missing MISTRAL_API_KEY: set it in the environment before running.
  • URL OCR fails: the URL likely is not publicly accessible; upload the file.
  • Large files: upload supports large files, but very large PDFs may need page selection (--pages) or batch processing.

References

  • API + parameters: references/mistral_ocr_api.md
  • Output mapping rules (placeholders to extracted images/tables): references/output_mapping.md
  • Example annotation prompts for common document types: references/annotation_prompts.md