Prism is a multimodal RAG (Retrieval-Augmented Generation) system that enables intelligent search and question-answering across documents (PDF, DOCX), images, and audio files. Built with a local-first approach, all processing happens on-device with no cloud dependencies, ensuring complete privacy and security.

• 🔍 Multimodal Search - Search across documents, images, and audio with natural language
• 🧠 Document Q&A - Ask questions about uploaded documents using Mistral 7B LLM
• 🏠 Local-First - All processing happens on your device, no cloud dependencies
• 📄 Document Processing - Support for PDF and DOCX files with intelligent chunking
• 🎨 Modern UI - Beautiful React interface with Tailwind CSS and Framer Motion
• 🚀 Fast API - High-performance FastAPI backend with async processing
• 🔒 Privacy-Focused - Your data never leaves your device

Implemented Features:

• ✅ Document upload and processing (PDF/DOCX)
• ✅ Intelligent text chunking with token awareness
• ✅ Document Q&A using Mistral 7B via llama.cpp
• ✅ Modern React frontend with multiple interfaces
• ✅ FastAPI backend with comprehensive endpoints
• ✅ Local LLM integration for offline operation

In Development:

• 🔨 Image OCR and processing
• 🔨 Audio transcription
• 🔨 Vector embeddings for semantic search
• 🔨 Full multimodal search interface

• Python 3.8+ with pip
• Node.js 16+ with npm
• 8GB+ RAM (for running Mistral 7B)
• 4GB+ disk space (for model storage)

git clone <repository-url>
cd prism

cd backend

# Create and activate virtual environment
python -m venv .venv
.venv\Scripts\activate  # Windows
# source .venv/bin/activate  # macOS/Linux

# Install dependencies
pip install -r requirements.txt

# Create models directory
mkdir -p models/llm

# Download Mistral 7B (4.1GB)
curl -L -o models/llm/mistral-7b-instruct-v0.2.Q4_K_M.gguf https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_K_M.gguf

cd frontend
npm install

Terminal 1 (Backend):

cd backend
.venv\Scripts\activate
python -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

Terminal 2 (Frontend):

cd frontend
npm run dev

Open your browser and go to http://localhost:3000 🎉

📋 For detailed setup instructions, see SETUP_QA.md

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   React Frontend │    │  FastAPI Backend │    │ Mistral 7B LLM  │
│   (Port 3000)   │◄──►│   (Port 8000)   │◄──►│  (llama.cpp)    │
└─────────────────┘    └─────────────────┘    └─────────────────┘
         │                       │                       │
         ▼                       ▼                       ▼
┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│  File Upload    │    │ Document        │    │ Question        │
│  Interface      │    │ Processing      │    │ Answering       │
└─────────────────┘    └─────────────────┘    └─────────────────┘

Frontend:

• React 18 with Vite
• Tailwind CSS for styling
• Framer Motion for animations
• Lucide React for icons
• Axios for API calls

Backend:

• FastAPI for high-performance APIs
• llama.cpp for LLM inference
• PyPDF2 & python-docx for document parsing
• tiktoken for intelligent text chunking

AI/ML:

• Mistral 7B Instruct v0.2 (GGUF format)
• Token-aware document chunking
• Local inference with no cloud dependencies

prism/
├── 🎨 frontend/                 # React application
│   ├── src/
│   │   ├── components/          # UI components
│   │   │   ├── DocumentQA.jsx   # Q&A interface
│   │   │   ├── FileUpload.jsx   # File upload UI
│   │   │   ├── SearchInterface.jsx
│   │   │   └── ...
│   │   ├── App.jsx              # Main app component
│   │   └── index.css            # Tailwind styles
│   └── package.json
│
├── 🔧 backend/                  # FastAPI application
│   ├── app/
│   │   ├── main.py              # API endpoints
│   │   └── services/
│   │       ├── llm_service.py   # Mistral LLM integration
│   │       └── qa_service.py    # Document Q&A logic
│   ├── ingestion/
│   │   ├── parse_pdf.py         # PDF/DOCX processing
│   │   └── chunker.py           # Intelligent text chunking
│   └── requirements.txt
│
├── 🧠 models/
│   └── llm/                     # LLM model storage
│       └── mistral-7b-instruct-v0.2.Q4_K_M.gguf
│
├── 📊 data/                     # Application data
│   ├── uploads/                 # Uploaded files
│   ├── processed/               # Processed documents
│   └── indices/                 # Search indices
│
└── 📚 docs/                     # Documentation
    ├── architecture.md
    └── runbook.md

1. Upload Documents - Upload PDF or DOCX files via the Q&A interface
2. Processing - Documents are automatically chunked and indexed
3. Ask Questions - Type natural language questions about your documents
4. Get Answers - Receive contextual answers with source citations

Example Questions:

• "What are the main conclusions in this report?"
• "Summarize the budget section"
• "What recommendations are mentioned?"
• "Find information about project timelines"

• Text Search - Search across all processed documents
• Multimodal Support - Future support for image and audio search
• Smart Results - Relevance scoring and source citations

Edit backend/app/services/llm_service.py:

# Context window size (adjust for memory usage)
self.n_ctx = 4096

# Thread count (adjust for CPU cores)
self.n_threads = 4

# Temperature (0.0 = focused, 1.0 = creative)
temperature = 0.7

Edit backend/ingestion/chunker.py:

# Chunk size in tokens
chunk_size = 1000

# Overlap between chunks
chunk_overlap = 200

cd frontend
npm run dev     # Start development server
npm run build   # Build for production
npm run lint    # Run ESLint

cd backend
.venv\Scripts\activate
python -m uvicorn app.main:app --reload  # Start with auto-reload

# Run backend tests
cd backend
python -m pytest tests/

# Run frontend tests
cd frontend
npm test

The FastAPI backend provides comprehensive API documentation:

• Interactive Docs: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

• Local Processing - All data processing happens on your device
• No Cloud Dependencies - Documents never leave your machine
• Offline Operation - Works completely offline after setup
• Data Control - You own and control all your data

We welcome contributions! Please read our contributing guidelines:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Documentation: Check SETUP_QA.md for detailed setup
• Issues: Report bugs via GitHub Issues
• Discussions: Join discussions for questions and ideas