Local AI Agent with RAG

Overview

This project implements a local AI agent capable of answering questions about pizza restaurant reviews. It utilizes the Retrieval-Augmented Generation (RAG) pattern, leveraging a local Large Language Model (LLM) via Ollama and a vector database (Chroma) to provide contextually relevant answers based on a provided dataset of reviews.

Purpose and Problem Solved

The primary purpose of this project is to demonstrate how to build a question-answering system using local AI models and RAG. It addresses the problem of efficiently querying and summarizing information from a collection of text documents (restaurant reviews in this case) without relying on external cloud services.

Key Features

• Local LLM Integration: Uses Ollama to run LLMs (like Llama 3.2) locally.
• Retrieval-Augmented Generation (RAG): Employs Langchain and Chroma to retrieve relevant review snippets before generating an answer.
• Vector Embeddings: Uses mxbai-embed-large via langchain-ollama for creating text embeddings.
• Persistent Vector Store: Creates and utilizes a persistent Chroma vector database (chrome_langchain_db/) to store review embeddings.
• CSV Data Source: Reads restaurant reviews from a realistic_restaurant_reviews.csv file.
• Interactive Q&A: Provides a command-line interface to ask questions about the reviews.

Requirements

• Python 3.x
• Ollama installed and running with a model like llama3.2 and mxbai-embed-large pulled.
• The following Python packages (install via pip install -r requirements.txt):
  • langchain
  • langchain-ollama
  • langchain-chroma
  • pandas

File Structure

. 
├── chrome_langchain_db/    # Chroma vector store directory (created on first run)
├── main.py                 # Main application script for interactive Q&A
├── realistic_restaurant_reviews.csv # Dataset containing restaurant reviews
├── requirements.txt        # Project dependencies
├── vector.py               # Script for embedding data and setting up the retriever
└── README.md               # This file

Getting Started

1. Clone the repository:

   git clone <repository-url>
   cd LocalAIAgentWithRAG-main

2. Ensure Ollama is running: Make sure the Ollama service is active and you have pulled the necessary models:

   ollama pull llama3.2
   ollama pull mxbai-embed-large

3. Install dependencies:

   pip install -r requirements.txt

4. Run the application: The first time you run main.py, it will execute vector.py to create the vector store. Subsequent runs will use the existing store.

   python main.py

5. Ask questions: Follow the prompts in the terminal to ask questions about the pizza reviews. Type q to quit.

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues to improve the project.

License

(Optional: Add your preferred license information here, e.g., MIT License)

Contact

• 🌐 Connect with me on LinkedIn: Syed Abdullah Shah
• 📩 Email: sa.abdullahshah.2001@gmail.com