LangChain Chroma Agent

This project demonstrates the use of LangChain with Chroma for document embedding and retrieval. It leverages Azure OpenAI for generating embeddings and executing chat-based interactions.

Features

• PDF Embedding: Convert PDF documents into embeddings using Azure OpenAI
• Chroma Vector Store: Store and retrieve document embeddings using Chroma DB
• Chunking: Intelligent document chunking with customizable size and overlap
• Chat Agent: Interact with the system using a chat-based interface powered by LangChain and Azure OpenAI

Prerequisites

• Python 3.8+
• Azure OpenAI account
• Chroma DB (running in Docker)

Setup

1. Clone the repository:

   git clone <repository-url>
   cd <repository-directory>

2. Start Chroma DB:

   docker run -d -p 8000:8000 -v C:/chroma/data:/vector_data -e CHROMA_SERVER_CORS_ALLOW_ORIGINS='["http://localhost:8090"]' -e PERSIST_DIRECTORY=/vector_data --name chromadb chromadb/chroma

3. Install dependencies:

   pip install -r requirements.txt

4. Environment Variables:

   • Copy .env_sample to .env and fill in the required API keys and endpoints.

   Environment Variables Description

   • OPENAI_API_KEY: Your OpenAI API key
   • AZURE_OPENAI_API_KEY: Your Azure OpenAI API key
   • AZURE_OPENAI_ENDPOINT: The endpoint URL for Azure OpenAI
   • AZURE_OPENAI_API_VERSION: The API version for Azure OpenAI
   • LANGSMITH_TRACING: Enable or disable LangSmith tracing (true/false)
   • LANGSMITH_ENDPOINT: The endpoint URL for LangSmith
   • LANGSMITH_API_KEY: Your LangSmith API key
   • LANGSMITH_PROJECT: The project name for LangSmith

5. Prepare PDF Embeddings:

   • Place your PDF files in the ./data/input/ directory
   • Run the embedding process to create and store document embeddings in Chroma DB

Usage

1. Start the Application:

   python agent.py

2. Interact with Documents:

   • The system will process PDFs from the input directory
   • Use the chat interface to ask questions about your documents
   • The agent will retrieve relevant information using the Chroma vector store

Configuration

You can customize the document processing by adjusting these parameters:

• Chunk size (default: 1000 characters)
• Chunk overlap (default: 200 characters)
• Collection name for vector storage

Project Structure

• agent.py: Main script to run the chat agent
• embeddings_tools.py: Functions to create and manage document embeddings with Chroma
• tools.py: Utility functions for file writing and message extraction
• .env_sample: Sample environment configuration file
• data/: Directory for input PDFs and output markdown files

Acknowledgments

• LangChain for the foundational framework
• Chroma for vector storage and retrieval
• Azure OpenAI for embedding and chat capabilities