ConstituteAI GPT is an AI-powered chatbot designed to provide accurate, simplified, and context-specific legal information from the Indian Constitution, related amendments, and case law. It serves a wide range of users including citizens, legal professionals, and law enforcement officers, offering intuitive and real-time access to legal knowledge.
- Constitutional Data Retrieval: Efficient querying of articles, amendments, and case laws with simplified explanations of complex legal texts.
- Custom Responses: Personalized responses using advanced AI models like Google Generative AI (Gemini), with context-aware explanations tailored to user needs.
- Advanced Search Engine: Uses HuggingFace models for embedding legal texts into vectors and FAISS for fast, scalable vector storage and retrieval.
- Natural Language Processing: Understands user intent and context for better query handling and simplifies legal concepts for non-legal audiences.
- Multifaceted Use Cases:
- Assists citizens in understanding their rights.
- Supports law enforcement with quick legal references.
- Enables in-depth research for legal professionals.
| Technology | Purpose |
|---|---|
| LangChain | Manages prompt handling and interaction flows |
| Google Generative AI | Generates context-aware and simplified responses |
| HuggingFace | Embeds constitutional text into vectors |
| FAISS | Enables efficient vector storage and retrieval |
| pypdf | Extracts content from legal PDF documents |
| re | Performs text pattern matching during preprocessing |
| FastAPI | Serves machine learning models through APIs |
| MongoDB | Stores user interactions and feedback data |
- ml-model branch: Contains the machine learning model code.
- Python 3.8+
- FAISS-cpu(Vector Database)
- Virtual environment tool (optional but recommended)
- Clone the repository:
git clone https://github.com/wahid18-maqs/ConstitutionGPT.git - Switch to the ml-model branch:
git checkout ml-model - Create and activate a virtual environment:
python -m venv venv
source venv/bin/activate(On Windows:venv\Scripts\activate) - Install dependencies:
pip install -r requirements.txt - Run the FastAPI server:
uvicorn main:app --reload
- Spearheaded development of a 97% accurate chatbot for Indian Constitution consultations.
- Built a Retrieval-Augmented Generation (RAG) pipeline to optimize model performance by storing vectorized data in FAISS using the all-MiniLM-L6-v2 embedding model via Hugging Face Inference API.
- Leveraged Gemini 1.5 Flash LLM for fast, context-aware conversational chatbot responses.
- Utilized FastAPI and React.js for interaction and query submission, with Docker ensuring deployment scalability.
- Data Collection: Extracts legal text from PDFs and organizes it into structured formats.
- Embedding and Storage: Converts text into vector embeddings using HuggingFace models and stores embeddings in FAISS for fast querying.
- User Interaction: Handles user queries via frontend app and queries the FastAPI backend for relevant legal information.
- Response Generation: Retrieves relevant legal articles and explanations, simplifying responses using Google Generative AI.
- POST /ask: Accepts user queries and returns legal insights.
- Citizens: Ask about rights and legal remedies.
- Lawyers: Query specific constitutional clauses and case laws.
- Law Enforcement: Obtain quick legal references for enforcement purposes.