BertrandConxy/Tax-Geek-AI-chatbot

This is an AI chatbot that leverages pre-trained AI models like gpt-4 as well as the Rwanda Tax documentations in order to answer any user query related to taxes. # Tax Geek - AI Chatbot This project is a LangChain-powered chatbot app designed to allow users to ask questions related to tax law. It leverages OpenAI embeddings to process and retrieve contextually relevant answers from a collection of tax law documents using a Retrieval-Augmented Generation (RAG) system. Additionally, LangSmith is used to monitor and debug every user interaction, and the system includes pytest tests for ensuring reliable performance. ## Features - **RAG System:** Combines retrieval from document embeddings and OpenAI's LLM to provide accurate and context-specific answers. - **Vector Embeddings:** Automatically processes and stores PDF files into embeddings using FAISS and OpenAI. - **Interactive UI:** User-friendly interface built with Streamlit for seamless interaction. - **LangSmith Monitoring:** Tracks and analyzes all interactions to improve the chatbot's performance and reliability. - **Test Coverage:** Pytest tests ensure the RAG system functions as expected and retrieves accurate results. ## Prerequisites Before running the application, ensure you have the following: 1. Python 3.8 or later 2. Required Python libraries: - `streamlit` - `faiss-cpu` - `pypdf` - `langchain` - `langchain-openai` - `pytest` - `langsmith` ## Folder Structure - **`./documentations`**: Place your tax-related PDF documents here. - **`app.py`**: Main application file. - **`create_vectordb.py`**: Creates vectordb with document embeddings and save the vectordb locally. - **`test_rag.py`**: Contains pytest tests for the RAG system. ## How to Run 1. Clone the repository: ```bash git clone github.com cd Tax-Geek-AI-chatbot ``` 2. Create virtual env ```bash python -m venv venv ``` 3. Install the dependencies using: ```bash pip install -r requirements.txt ``` 4. Ensure tax-related PDF documents are in the `documentations` folder. 5. Create `.env` file for credentials ``` OPENAI_API_KEY LANGSMITH_TRACING=True LANGSMI …