Skip to content

README.md

Latest commit

 

History

History
124 lines (97 loc) · 3.84 KB

README.md

File metadata and controls

124 lines (97 loc) · 3.84 KB

📚 Simple RAG System with LangChain and Ollama

This repository contains a basic Retrieval-Augmented Generation (RAG) system built using LangChain, ChromaDB, and Ollama. The goal is to demonstrate how to use a Large Language Model (LLM) to answer questions based on a custom knowledge base (PDF documents) running entirely locally.

✨ Features

  • Custom Knowledge Base: Load your own PDF documents to create a domain-specific knowledge base.
  • Local LLM and Embeddings: Utilizes Ollama to run both the embedding model (nomic-embed-text) and the Large Language Model (mistral) locally on your machine, ensuring data privacy and reducing API costs.
  • Vector Database: Employs ChromaDB as a lightweight, local vector store to efficiently search and retrieve relevant document chunks.
  • Modular Design: Separated concerns for embedding function, database population, and querying.
  • Makefile for Convenience: Simple commands to set up the database and run the query agent.

🚀 Getting Started

Follow these steps to get your RAG system up and running.

Prerequisites

Before you begin, ensure you have the following installed:

  1. Python 3.8+: 
    python3 --version
  2. pip (Python package installer): 
    pip --version
  3. Ollama: Download and install Ollama from ollama.com.
    • Once installed, start the Ollama server in a dedicated terminal: 
      ollama serve
    • Pull the necessary models: 
      ollama pull nomic-embed-text
ollama pull mistral

Installation

  1. Clone the Repository:

    git clone https://github.com/callmemehdy/docAssistantChatBot.git
cd docAssistantChatBot

  2. Create a Virtual Environment (Recommended):

    python3 -m venv v_env
source v_env/bin/activate  # On Windows, use `v_env\Scripts\activate`

  3. Install Python Dependencies:

    pip install -r requirements.txt

Prepare Your Data

  1. Create a folder named data in the root of your project directory: 
    mkdir data
  2. Place your PDF documents (.pdf files) that you want to query into this data folder. For example: 
    callmemehdy-docAssistantChatBot/
├── data/
│   ├── my_document_1.pdf
│   └── my_document_2.pdf
└── ...

Populate the Database

This step processes your PDF documents, splits them into manageable chunks, creates numerical embeddings for these chunks using nomic-embed-text, and stores them in a local ChromaDB instance.

  • Using Makefile (Recommended):

    make setup_db

    This will create a chroma directory in your project containing the vector database.

  • Manually:

    python3 ./src/pop_db.py

  • To reset and rebuild the database:

    make reset_db
# or manually:
python3 ./src/pop_db.py --reset

💬 Querying the System

Once the database is populated, you can start asking questions! The system will retrieve relevant information from your PDFs and use the mistral LLM to formulate an answer.

  • Using Makefile (Recommended):

    make all

    This will first ensure the database is set up, then start the interactive query agent.

  • Manually:

    python3 ./src/evaluate_queries.py

The agent will prompt you with query:. Type your question and press Enter. To exit, type EXIT.

Example Interaction

In case you feed the agent the MariaDB documentation PDF:

agent loading...
query: which port the mariadb service runs on?
you will get your response after a while...
Response: port 3306.
query: EXIT
Bye!