Retrieval-Augmented Generation (RAG) System for Legal Document Q&A This project demonstrates how to build a Retrieval-Augmented Generation (RAG) pipeline that can answer questions about legal documents and retrieve relevant clauses using open-source tools like LangChain, FAISS, and Hugging Face Transformers.

🔍 Use case: Automating legal document understanding — including contract clause retrieval, regulatory Q&A, and compliance insights.

🚀 Features ✅ Load and process legal contract datasets.

📄 Intelligent document chunking using RecursiveCharacterTextSplitter.

🔍 Semantic search via dense embeddings and FAISS vector store.

🤖 Query answering using LLMs like Mistral-7B or other Hugging Face-hosted models.

📚 End-to-end RAG pipeline using LangChain.

⚙️ Modular and extendable — plug in your own datasets, models, or prompts.

🧠 Technologies Used LangChain – for chaining retrieval and LLM generation.

SentenceTransformers – for generating text embeddings (MiniLM-L6-v2).

FAISS – for fast similarity search over document chunks.

Transformers (Hugging Face) – to load and run LLMs.

BitsAndBytes – for 4-bit quantized LLM loading.

Google Colab – for development and GPU experimentation.

Folder structure: . ├── RAG_project3.ipynb # Jupyter Notebook with full pipeline ├── data/ # (optional) directory for PDF or text contracts ├── README.md # You are here!