doc7

Why we included this project

doc7 is a single CLI tool that turns PDFs, Office files, scans, and screenshots into Markdown by sending each page to an OpenAI-compatible multimodal model you already run, whether that is a local vision model in LM Studio or Ollama or a private remote endpoint. There is no bundled OCR stack and no dependency on a paid document-parsing service, so sensitive documents never leave your own hardware and costs stay predictable. The project handles raster-only pages well, recovering equations and figures, including the relationships inside diagrams that plain text extraction would miss. It also ships batch processing, an HTTP service, and a Go SDK alongside the binary. Teams building knowledge bases or RAG indexes, and anyone normalizing internal document workflows, get a self-contained way to normalize messy inputs into text their models can actually search and quote.