LegalClassifier

AI-assisted legal document analysis and procedural roadmap generation built with React and FastAPI.

This project was created for the Google Gen AI Exchange Hackathon. It helps users:

• upload legal documents in pdf, docx, or txt format
• extract plain text, including OCR fallback for scanned PDF pages
• generate structured summaries with alerts and key clauses
• create step-by-step procedural roadmaps from natural-language goals
• ask follow-up questions about the current document or roadmap
• export the current result view as a PDF

Live demo: legalclassifier.netlify.app

Current Scope

The app is an MVP focused on a few clear workflows:

• document analysis for one or more uploaded files
• roadmap generation for legal or procedural goals
• multilingual output in English, Tamil, Hindi, and Malayalam
• follow-up Q and A using the currently selected result as context

What it does not include yet:

• user accounts or saved cloud history
• a vector database or cross-session memory
• document comparison across multiple files
• verified legal advice or jurisdiction-specific guarantees

Tech Stack

• Frontend: React, Vite, Axios, Tailwind CSS
• Backend: FastAPI, PyMuPDF, python-docx, Mammoth, Pillow, Tesseract OCR
• AI: Gemini via google-generativeai with gemini-2.5-flash
• Deployment: Netlify for the frontend, Render-compatible FastAPI backend

Running Locally

Backend

cd backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload

Create backend/.env with:

GOOGLE_API_KEY=your_api_key_here

Frontend

cd frontend
npm install
npm run dev

Optional frontend environment variable:

VITE_API_URL=http://localhost:8000

Notes

• OCR support is currently implemented for scanned PDF pages through Tesseract.
• AI output is generated from prompts and may be incomplete or wrong. Treat it as assistance, not legal advice.
• The current PDF export captures the visible result view.

Future Improvements

• persistent user history
• stronger multi-document comparison and synthesis
• paginated PDF export for long reports
• backend tests and structured logging
• better chunking and code-splitting for the frontend bundle

License

Apache 2.0. See LICENSE.