Team: Crewmate | Live Demo | Demo Video
Built for the Gen AI Exchange Hackathon
Generative AI for Demystifying Legal Documents
Legal SahAI is an AI-powered platform designed to make legal documents and workflows understandable, accessible, and efficient.
It bridges the gap between citizens and legal professionals using Generative AI, RAG, and multimodal processing.
Upload any legal document — PDF, DOCX, image, or handwritten note — and unlock instant insights:
- Chat with PDF: Conversationally query your document.
- AI Summary: Concise, plain-language summary of key points.
- Timeline Extraction: Automatically organizes key events and dates.
- Suggested Questions: Generates relevant legal queries.
- Predictive Analysis: Forecasts potential case outcomes.
- Case Law Suggestions: Retrieves supporting precedents via Kanoon API.
- Smart Risk Audit: Visually highlights risky clauses and explains their compliance status under specific Indian laws.
A secure bridge between Help Seekers and Legal Professionals:
- For Clients: Search, filter, and contact verified lawyers.
- For Lawyers: Manage client requests and chat securely in-platform.
Guides users in filling complex government or legal forms with:
- Smart field-by-field hints and data suggestions.
- Automated recognition of form blanks and required fields.
Ask general law-related questions — get instant, multilingual, AI-powered legal guidance.
Our platform is built on a robust, service-oriented architecture designed for scalability and advanced AI processing.
- Conditional RAG (Retrieval-Augmented Generation): Our system is more than a simple RAG. A Router Agent intelligently decides the best source for an answer:
- Pinecone (Vector DB): For information inside the user's private documents.
- KanoonAPI (External API): For historical legal precedents and case law.
- Google Custom Search: For general web-based legal queries.
- True Multimodality (Input & Ingestion):
- Interaction: Users can ask questions via typed text or voice (using Google Speech-to-Text).
- Ingestion: The system can process PDFs, DOCX, images, and handwritten notes (using Google Cloud Vision).
- "Translate -> Process -> Translate Back" Architecture: We use Google Cloud Translation to translate all non-English queries into English for high-accuracy processing by Gemini, then translate the answer back, ensuring a seamless multilingual experience.
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | React.js, Tailwind CSS, Zustand |
Client-side UI, state management, and user experience. |
| Backend (User Service) | Node.js, Express.js |
Manages user auth, profiles, and lawyer-client chat. |
| AI Core | Python, FastAPI |
Handles all AI processing, RAG, and document ingestion. |
| Databases | MongoDB, Pinecone |
MongoDB for user data/chats. Pinecone for vector storage. |
| Cloud & AI Services | Google Cloud (Gemini, Vision, Translate, Speech-to-Text, Search, Storage) |
Multimodal and multilingual AI backbone |
Built using open APIs and tools from:
Google Cloud(Gemini, Vision, Translate, STT, Search, Storage)KanoonAPIfor legal case dataLangChainfor AI orchestration
“Empowering citizens with AI-driven legal clarity.”