Many lawyers and legal tech teams struggle with the same issue: they need to search and query sensitive client documents and transcripts, but uploading privileged material to OpenAI, Anthropic, or Gemini is a non-starter for ethics and confidentiality. That’s why I built a fully self-hosted tool using local open-source LLMs (Ollama) with mechanically verified page citations for them. Every answer can be checked against the exact text span no black-box hallucinations. It started as a clean RAG pipeline and grew into a complete document intelligence stack: matter isolation, OCR, table extraction, review grids, and clause checklists. Deliberately not an “AI lawyer” — just a transparent retrieval assistant that tells you what your documents actually say (or honestly admits when it can’t). Happy to answer questions on architecture, deployment, or customization. Feedback and contributions welcome!

Legal Document Chat for Attorneys, open

Confidential legal document chat, fully local + cited

Confidential legal document chat, fully local + cited