Contract-Advisor-RAG-Towards-Building-A-High-Precision-Legal-Expert-LLM-APP

This repository contains the codebase for building, evaluating, and improving a Retrieval Augmented Generation (RAG) system for Contract Q&A. The objective is to develop a powerful contract assistant and eventually a fully autonomous contract bot.

Project Overview

The project aims to leverage RAG technology to create an AI-powered contract advisor capable of answering questions related to legal contracts. By combining large language models with external data sources, the system aims to provide accurate and context-rich responses to user queries. The development process involves:

• Researching ways to improve RAG systems, including efficiency, scalability, personalization, contextualization, and bias reduction.
• Building a simple Q&A pipeline with RAG using Langchain, a leading LLM application framework.
• Creating a RAG evaluation pipeline with RAGAS to assess the performance of the system.
• Ideating and implementing enhancements to optimize the system for Contract Q&A.
• Interpretation & Reporting: Presenting the outcomes of the project with a short deck and providing performance metrics to quantify incremental improvements.

Folder Structure

• backend/: Contains the Flask backend for serving the RAG model and handling API requests.
• frontend/: Houses the React frontend for user interaction and displaying contract Q&A results.
• logs/: Stores log files for backend and frontend activities.
• notebooks/: Includes Jupyter notebooks for data exploration, model training, and performance evaluation.
• prompts/: Stores prompt templates and examples for RAG model input.
• scripts/: Contains utility scripts for data preprocessing, model training, and evaluation.
• tests/: Holds unit tests and integration tests for backend and frontend functionalities.
• workflows/: Contains GitHub Actions workflows for automated testing, linting, and deployment.
• .env_example: Example environment variables file for configuring the Flask backend.
• .gitignore: Specifies intentionally untracked files to ignore in version control.
• Makefile: Provides convenient commands for setting up, running tests, and other project tasks.
• README.md: Main repository documentation providing an overview of the project and instructions for setup.

Setup Instructions

Backend Setup:

• Navigate to the backend/ directory.
• Create a virtual environment: python -m venv venv.
• Activate the virtual environment: source venv/bin/activate.
• Install dependencies: pip install -r requirements.txt.
• Copy .env_example to .env and configure environment variables.
• Run the Flask app: flask run.

Frontend Setup:

• Navigate to the frontend/ directory.
• Install dependencies: npm install.
• Start the React app: npm start.

Testing:

• Run backend tests: pytest backend/tests.
• Run frontend tests: npm test.

Deployment:

• Utilize GitHub Actions workflows for CI/CD pipeline setup.

License

This project is licensed under the MIT License - see the LICENSE file for details.