PrivateGPT
December 5, 2024

PrivateGPT

AIPrivacyDocumentsRAGPython

Chat with your documents privately. No data leaves your machine. Ask questions about PDFs, code, and more.

VIEW ON GITHUB

PrivateGPT lets you chat with your documents using local LLMs. 100% private - your data never leaves your computer.

Quick Deploy

```bash # Clone the repo git clone https://github.com/imartinez/privateGPT.git cd privateGPT

# Install dependencies pip install -r requirements.txt

# Download a model (e.g., Llama 2) # Place in models/ folder

# Ingest your documents python ingest.py

# Start chatting python privateGPT.py ```

Docker Deploy

```bash docker-compose up -d # Access at http://localhost:8080 ```

Use Cases

Legal Document Review: Ask questions about contracts without uploading sensitive data to the cloud.

Research Assistant: Chat with academic papers, extract key findings.

Codebase Q&A: Ingest your code, ask "how does the auth system work?"

Company Knowledge Base: Build internal chatbots on proprietary docs.

Medical Records: Query patient data privately (HIPAA compliant).

Financial Analysis: Analyze reports without data exposure.

Supported Formats

- PDF, DOCX, TXT - CSV, XLSX - Python, JavaScript, and other code files - Markdown, HTML - Email archives

Pro Tips

- Use better models for better answers - Chunk size affects retrieval quality - More documents = slower but more comprehensive - GPU dramatically speeds up responses

Hardware

- CPU-only: Works but slow - 8GB VRAM: Good performance - 16GB+ VRAM: Best experience