Create a private QnA chatbot on your documents without relying on the internet. Leverage the power of local LLMs for complete privacy and security - none of your data ever leaves your local environment.
Inspired by imartinez/privateGPT
- 100% Private - All processing happens locally. Your documents never leave your machine
- No Internet Required - Works completely offline after initial setup
- Multiple Document Formats - Support for PDF, TXT, DOC, DOCX, and more
- Local LLM Support - Uses GPT4All for local language model inference
- Interactive UI - Clean web interface for document upload and chat
- Document Ingestion - Automatic chunking and embedding of your documents
- Source Citations - See which parts of your documents informed each answer
- Python 3.8 or later
- Node.js v18.12.1 or later
- Minimum 16GB RAM
-
Clone the repository
git clone https://github.com/SamurAIGPT/EmbedAI.git cd EmbedAI -
Start the client
cd client npm install npm run dev -
Start the server (in a new terminal)
cd server pip install -r requirements.txt python privateGPT.py -
Open the app
Navigate to http://localhost:3000 and click "Download Model" to get the required LLM.
EmbedAI/
├── client/ # Next.js frontend
│ ├── components/ # React components
│ └── pages/ # App pages
└── server/ # Python backend
├── privateGPT.py # Main server
└── ingest.py # Document processing
- Upload Documents - Drop your files into the web interface
- Automatic Processing - Documents are chunked and embedded locally
- Ask Questions - Chat naturally with your document corpus
- Get Answers - Receive responses with source citations
| Format | Extension |
|---|---|
.pdf |
|
| Text | .txt |
| Word | .doc, .docx |
| Markdown | .md |
The default model is GPT4All. You can configure:
- Model selection
- Chunk size for document processing
- Number of source documents to retrieve
Model download fails?
- Ensure you have a stable internet connection for the initial model download
- Check that you have enough disk space (~4GB for the model)
Slow responses?
- Local LLM inference requires significant CPU/RAM
- Consider using a machine with at least 16GB RAM
Context window errors?
- Try reducing the document chunk size
- Split large documents into smaller files
Contributions are welcome! Please feel free to submit a Pull Request.
- Open-Custom-GPT - Build Custom GPTs with the Assistants API
- Text-To-Video-AI - Generate videos from text
MIT License - see LICENSE for details.