DONNA - Domain-Specific Legal AI Assistant

"AI is transforming industries — but law remains locked in complexity. DONNA is built to change that."

DONNA (Document-Oriented Legal AI Assistant) is a legal AI system designed to make legal documents and processes accessible, understandable, and efficient. It combines state-of-the-art LLMs with intelligent retrieval pipelines to assist lawyers, students, and citizens in navigating complex legal information.

---

Features

• Multi-LLM Architecture
  • Groq Llama 3.1 for precision legal reasoning
  • Google Gemini Flash for structured summarization
• RAG-Powered Pipeline
  • Smart chunking + semantic retrieval (Chroma + HuggingFace embeddings)
• Human-Centered Design
  • Session-based chat (temporary memory, like a legal co-pilot)
  • Inline news digest (no distraction, no extra clicks)
• Efficiency First
  • Embedding caching for fast re-queries
  • GPU/CPU-optimized batching

---

Tech Stack

• Backend: FastAPI + LangChain + ChromaDB + HuggingFace Embeddings
• LLMs: Groq Llama 3.1 (Q&A), Gemini 2.0 Flash (Summarization)
• Frontend: React.js + TailwindCSS
• OCR/Parsing: PyPDF2, pdf2image, pytesseract

---

Installation & Setup

Prerequisites

• Python 3.10+
• Node.js 18+
• Tesseract OCR installed on your system
• Poppler (required for pdf2image)
• API Keys for:
  • Groq
  • Google Gemini
  • News API

---

Environment Variables

1. Copy the template file:

cp .env.example .env

2. Replace the placeholder values in .env with your actual API keys:

# .env
GROQ_API_KEY="your_real_groq_api_key"
GEMINI_API_KEY="your_real_gemini_api_key"
NEWS_API_KEY="your_real_news_api_key"

⚠️ Do not commit your .env file to GitHub, it is already ignored in .gitignore.

Make sure to setup backend and frontend in different terminals. (use cmd, not powershell)

3. Backend Setup

cd backend
pip install -r requirements.txt
uvicorn main:app --reload --port 8000

4. Frontend Setup

cd frontend
npm install
npm start

---

Project structure

DONNA-LegalAI/
├── .env              # API keys (not committed, you will be creating this file.)
├── .env.example      # Example template for ".env"
├── .gitignore        # Ignore env
├── LICENSE           # MIT License
├── README.md         # Project documentation
├── backend/          # Python FastAPI backend
│   ├── main.py
│   └── src/
│       ├── __init__.py
│       ├── pdf_handler.py
│       ├── summarization_model.py
│       ├── qa_model.py
│       └── news_fetcher.py
└── frontend/         # React frontend
    ├── package.json
    ├── public/
    │   └──index.html
    └── src/
        ├── App.css
        ├── App.js
        ├── index.css
        ├── index.js
        ├──LandingPage.css
        └──LandingPage.js

---

Example Use Cases

• Law Students: Quickly summarize 200-page case law before class.
• Lawyers: Ask direct questions from a client's contract PDF.
• Policy Analysts: Track global legal reforms in real time.
• Citizens: Understand your rights without needing to parse jargon.

---

Research Basis

• Attention Is All You Need - Vaswani et al. (2017)
• Retrieval-Augmented Generation for Knowledge-Intensive NLP - Lewis et al. (2020)
• Recent advancements in domain-specific LLMs & efficient retrieval pipelines

---

License

This project is licensed under the MIT License.

---

Authors

• @Hrishhii