Offgrid PDF

Offgrid PDF is a 100% locally hosted PDF chat application that allows you to have interactive conversations with your PDF documents. Upload your PDF and start asking questions about its content - all while keeping your data completely private on your local machine.

Features

• 100% Local Processing: All data stays on your machine - no cloud services involved
• Privacy-Focused: Your documents are never sent to external servers
• Powered by Ollama: Uses local Large Language Models for document analysis and responses
• Intelligent PDF Analysis: Processes and understands complex documents
• Interactive Chat Interface: Have natural conversations about your document content
• Dark Mode UI: Easy on the eyes for extended reading sessions
• File-based Storage: Simple data handling without complex databases
• Fast Document Processing: Efficient chunking and vectorization of documents
• RAG Implementation: Retrieval-Augmented Generation for accurate, context-aware responses

System Requirements

• Python 3.8+
• Node.js 14+
• Ollama (for running local language models)
• 8GB+ RAM recommended
• 10GB+ free disk space (depending on the LLMs you use)
• GPU (optional) for faster model inference

Setup Instructions

1. Clone the Repository

git clone https://github.com/chandraprvkvsh/Offgrid-PDF.git
cd Offgrid-PDF

2. Install Ollama

Ollama is required to run the language models locally. Follow the installation instructions for your platform:

• Ollama Installation Guide

3. Download Required Models

OffgridPDF requires two models - a language model for chat and an embedding model for document processing. You can use any models offered by Ollama by changing the configuration in the backend (app/config.py).

# Download the language model
ollama pull llama3.2

# Download the embedding model
ollama pull mxbai-embed-large

# Start the Ollama service
ollama serve

4. Choose Your Setup Method

You can run Offgrid PDF in two ways:

Option A: Using Docker Compose (Recommended)

This is the simplest way to get started with minimal setup:

1. Make sure you have Docker and Docker Compose installed on your system
2. Run the application with:

docker-compose up

3. Access the application at http://localhost:3000

Note: You still need to install and run Ollama separately as shown in step 2 and 3.

Option B: Manual Setup

If you prefer to run the services without Docker, follow these steps:

Set Up Backend

Create and activate a virtual environment:

# Create a virtual environment
python -m venv venv

# Activate the virtual environment (Windows)
venv\Scripts\activate

# Activate the virtual environment (macOS/Linux)
source venv/bin/activate

Install the required Python packages:

pip install -r requirements.txt

Set Up Frontend

Navigate to the frontend directory and install the required packages:

cd frontend
npm install

If you encounter peer dependency issues, you can use:

npm install --legacy-peer-deps

Running the Application

Option A: Using Docker Compose

Start both the frontend and backend with a single command:

docker-compose up

This will build and start both services. To run in detached mode:

docker-compose up -d

To stop the services:

docker-compose down

Option B: Manual Startup

1. Start the Backend Server

Make sure your virtual environment is activated, then start the backend server:

# From the root project directory
uvicorn backend.app.main:app --reload --port 8000

2. Start the Frontend Development Server

In a new terminal, navigate to the frontend directory and start the React development server:

cd frontend
npm start

This will start the frontend on http://localhost:3000

3. Access the Application

Open your web browser and navigate to:

http://localhost:3000

Usage

1. Upload a PDF: On the home page, drag and drop a PDF or click to select one
2. Wait for Processing: The system will process and analyze your document
3. Start Chatting: Ask questions about your document in the chat interface
4. View Responses: Get AI-generated answers based on your document content
5. Start New Chats: Use the "New Chat" button to start fresh conversations
6. Upload New Documents: Choose "Upload New PDF" to work with different documents

Troubleshooting

• Ollama Connection Issues: Ensure Ollama is running with ollama serve
• Missing Models: Verify models are downloaded with ollama list
• Backend Errors: Check terminal output for Python errors
• Frontend Issues: Inspect browser console for JavaScript errors
• PDF Upload Failures: Ensure PDFs aren't password protected or corrupted

Architecture Overview

Offgrid PDF uses a modern web architecture:

• Backend: FastAPI Python application handling PDF processing and AI interactions
• Frontend: React-based single-page application with TypeScript
• Document Processing: LangChain for PDF parsing and chunking
• Vector Storage: FAISS for efficient document embeddings and retrieval
• LLM Integration: Direct integration with Ollama for local AI inference

License

MIT License

Acknowledgements

• Ollama for local LLM capabilities
• LangChain for document processing
• FastAPI for the backend framework
• React for the frontend framework