Document chat

🚀 Document Chat v0.1.0

Document Chat is a fully local, privacy-first document chat application that lets you upload files and interact with them using a locally running LLM — no cloud, no data leakage.

This is the first public release.

✨ Features

📄 Upload and chat with documents (PDF, text, code, images*)

🧠 Local LLM inference (no external APIs)

🔍 Intelligent document chunking & retrieval (RAG)

⚡ Fast semantic search with vector indexing

🔐 Fully offline & privacy-preserving

🖥️ Native desktop app (macOS / Windows / Linux)

🍎 Optimized for Apple Silicon (Metal backend)

• Image OCR support depends on local OCR tools (see notes below).

🧱 Architecture

Frontend: Electron + React (Vite)

Backend: Python (FastAPI)

Inference: llama.cpp (Metal / CPU)

Embeddings: SentenceTransformers

Vector Store: Local (FAISS / fallback)

Model Handling: On-demand download (models are not bundled)

📦 Download

Choose the binary for your platform from the assets below:

macOS (Apple Silicon): .dmg / .zip

Windows: .exe

Linux: .AppImage / .deb

⚠️ On first launch, the app will download the selected model automatically.

⚠️ Notes & Known Limitations

The first startup may take longer due to model download

Large documents are truncated to fit the model context window

OCR quality depends on local tools (Tesseract recommended)

App is unsigned (macOS may show a security warning)

🔐 Privacy

All processing happens locally on your machine.
No documents, embeddings, or prompts are sent to any server.

🛠️ Troubleshooting

If macOS blocks the app:
System Settings → Privacy & Security → Open Anyway

If OCR fails, install:

brew install tesseract

🙌 Feedback & Contributions

Issues, bug reports, and feature requests are welcome.
This project is under active development.