Repolode
Submit
magicrew

magicrew/doc7

Converts PDFs, Office files, scans, screenshots and diagrams into Markdown by sending pages to your own OpenAI-compatible vision model.

Open-source alternative to:MathpixABBYY FineReaderAdobe Acrobat

View on GitHub
Stars
721
Forks
38
Open issues
1
Last commit
2 months ago

What is it?

doc7 is a Go command-line tool and service that turns documents into Markdown for AI use. Each page is rendered and sent to a multimodal model, so text, tables, formulas, charts, diagrams and screenshots are read together rather than only extracted as characters. It does not require a separate OCR stack, and the model is whichever OpenAI-compatible endpoint you configure, such as LM Studio or Ollama running locally, or a remote provider.

Beyond the CLI, the same binary provides batch processing of directories, page-level retries, an asynchronous HTTP service with bearer-token auth, an MCP server for AI tools, and a Go SDK. Output includes merged Markdown, per-page Markdown, rendered images and a manifest with page provenance. Native text and data formats such as CSV, JSON and Markdown are converted locally without a model.

LibreOffice is needed for Office and OpenDocument files, and Chrome or Chromium is needed for HTML, EPUB, email and notebook rendering. Remote documents are sent to the configured endpoint, so the project suggests a local or private model for sensitive files.

Key features

  • 01Reads each page as an image through an OpenAI-compatible multimodal model, covering text, tables, formulas and charts.
  • 02Supports PDF, Office, OpenDocument, EPUB, email, Jupyter notebooks, common image formats and web pages.
  • 03Retries only failed pages with --resume and supports page-range selection.
  • 04Provides a CLI, an async HTTP service, an MCP server and a Go SDK from one binary.
  • 05Outputs merged Markdown, per-page files, page images and a manifest with provenance.

Use cases

  • Building a RAG or knowledge-base corpus from scanned reports and PDFs.
  • Making visual content such as charts and slide decks searchable by AI agents.
  • Converting sensitive documents with a private model that never leaves your infrastructure.
  • Batch-converting large document directories in CI or on a server.

Topics

#document-ai#pdf-to-markdown#vision-language-model#multimodal#ocr-alternative#markdown#rag#local-ai#docx-to-markdown#open-source

FAQ

What is doc7?
Converts PDFs, Office files, scans, screenshots and diagrams into Markdown by sending pages to your own OpenAI-compatible vision model. doc7 is a Go command-line tool and service that turns documents into Markdown for AI use. Each page is rendered and sent to a multimodal model, so text, tables, formulas, charts, diagrams and screenshots are read together rather than only extracted as characters. It does not require a separate OCR stack, and the model is whichever OpenAI-compatible endpoint you configure, such as LM Studio or Ollama running locally, or a remote provider.
Is doc7 free to use? What license does it have?
doc7 is open source under the MIT license. Check the license terms before using it in commercial products.
Is doc7 still maintained?
Updated within 6 months. The last commit to doc7 was on Aug 7, 2026.
What can I use doc7 for?
Building a RAG or knowledge-base corpus from scanned reports and PDFs. Making visual content such as charts and slide decks searchable by AI agents. Converting sensitive documents with a private model that never leaves your infrastructure. Batch-converting large document directories in CI or on a server.
Where can I get doc7?
The source code and installation instructions are on GitHub: https://github.com/magicrew/doc7

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects