GenAI + RAG: Business-Context-Aware Customer Assistant for Organic Farm-to-Table Business

Getting Started

This project aims to create a business context-aware customer assistant using Generative AI and Retrieval-Augmented Generation (RAG) techniques for Desi Bazar Agro. The assistant is built using Python, Flask, LangChain, ChromaDB, and OpenAI API.

Prerequisites

• Python installed

Cloning the Repository

Command to clone the repository:

git clone https://github.com/RyhanSunny/GenAI-RAG-Business-Context-Aware-Customer-Assistant-for-Desi-Bazar-Agro.git

Setting up a Virtual Environment

• In the root directory run:

python -m venv venv

• Activate the virtual environment:

source venv/bin/activate

• For Windows:

venv\Scripts\activate

• Install the required dependencies:

pip install -r requirements.txt

Running the Application

To run the application, make sure the virtual environment is active and run the following command:

python run.py

Docker

The easiest way to run the application is by using Docker:

docker-compose up

Process Flow

Tech Stack

• Python: Main programming language
• Flask: Web framework for API
• LangChain: Framework for LLM applications
• RetrievalQA: Implements RAG pattern
• PromptTemplate: Structures prompts
• ChromaDB: Vector database for embeddings; embedded instance of DuckDB
• OpenAI API:
  • ChatGPT model: Generates responses
  • Embeddings: Creates vector representations
• python-dotenv: Manages environment variables
• JSON: For configuration and API responses

RAG Process Flow

1. Text to embeddings (OpenAI)
2. Store in ChromaDB
3. Retrieve similar docs for query
4. LangChain combines docs + query
5. ChatGPT generates response
6. Flask serves response

Frontend

• HTML, CSS, and JavaScript: (index.html, style.css, chat.js)

Backend

• Python with Flask: (backend_api.py, app.py, website.py, run.py)
• Configuration: (config.json, requirements.txt)
• Docker setup: (Dockerfile, docker-compose.yml)
• Additional utilities: (highlight.min.js, highlightjs-copy.min.js, icons.js)

Key Scripts

• backend_api.py: Loads environment variables and configuration, initializes the RAG chatbot, and sets up Flask routes.
• rag_chatbot.py: Initializes the ChromaDB client, sets up LangChain components, and provides methods for adding documents and querying.
• populate_db.py: Populates ChromaDB with initial data about Desi Bazar Agro.

How to Run

1. Install dependencies:

   pip install -r requirements.txt
   

2. Populate the database:

   python populate_db.py
   

3. Run the backend API:

   python backend_api.py
   

Docker Setup

1. Build and run the Docker containers:

   docker-compose up
   

Health Check

Check the health of the application:

http://192.168.0.245:1338/health

Additional Notes

• To clear and repopulate the database:

  python populate_db.py --clear
  

• To add/update without clearing:

  python populate_db.py
  

• view_documents.py is a diagnostic tool to verify the contents of the database.

Issues and Improvements

• Context Awareness: Improve recognition of new conversation starters.
• Response Templates: Ensure the chatbot uses the exact response templates specified in the system prompt.
• Product Information: Ensure access to specific product information.
• Limited Context: Improve the context being used for responses.

TODO

• Create a system prompt that includes:
  • Language guidelines
  • Customer interaction guidelines
  • Instructions for handling common inquiries (location, delivery fee, price list, general delivery)
  • Invoice creation process
• Implement language detection and response formatting in the chatbot's code.
• Store the following in your vector database:
  • Product information and pricing
  • Delivery charges and areas
  • Business overview and contact information
• Create functions or logic in your chatbot to handle:
  • Formatting and sending invoices
  • Sharing the WhatsApp catalog link
  • Detecting and appropriately responding to "hm" as "how much"

Feel free to contribute to this project by forking the repository and creating pull requests. Your contributions are welcome!

MIT License