Chatbot with LangChain, RAG, and Django API

This project implements an AI chatbot leveraging LangChain, Retrieval-Augmented Generation (RAG), and Django to expose the chatbot's functionalities through API endpoints. The chatbot is designed to provide detailed and accurate responses based on structured data stored in a ChromaDB vector database.

Table of Contents

1. Features
2. Tech Stack
3. Installation
4. Usage
5. API Documentation
6. Directory Structure
7. Contributing
8. License

---

Features

• Conversational AI: Provides intelligent and context-aware responses based on Manu's resume and personality data.
• RAG-Based Search: Retrieves relevant information using similarity search from vectorized documents stored in ChromaDB.
• Agent Execution: Uses LangChain's reactive agent architecture for dynamic and task-specific responses.
• Django API: Exposes chatbot interactions through a RESTful API.

---

Tech Stack

• Backend: Python, Django
• AI Framework: LangChain, LangGraph
• Vector Database: ChromaDB
• Embedding Models: OpenAI GPT models
• Document Processing: LangChain Text Splitter, Docx2txt, PyPDFLoader

---

Installation

Prerequisites

1. Python 3.8+
2. Virtual environment tools like venv or conda
3. OpenAI API Key (store in a .env file)
4. Django and required libraries (requirements.txt file provided)

Steps

1. Clone the repository:

   git clone https://github.com/manugeorge04/PersonaAI.git
   cd PersonaAI

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Set up the data directory:

   • Inside the project root directory, create a data folder with the following structure:

     data/
     ├── Resume_Data/        # Store resume-related .docx files here
     └── Personality_Data/   # Store personality-related .docx files here
     

   • Add .docx files for:

     • Resume Data: Include one or more .docx files with professional experience, projects, and achievements.
     • Personality Data: Include .docx files with personal traits, hobbies, or character-related information.

   • Example:

     data/
     ├── Resume_Data/
     │   ├── resume1.docx
     │   ├── resume2.docx
     └── Personality_Data/
         ├── personality_traits.docx
         ├── hobbies.docx
     

   • You can add multiple .docx files to each folder as needed.

5. Set up the environment variables:

   • Create a .env file in the root directory with the following content:

     OPENAI_API_KEY=your_openai_api_key
     

6. Initialize the vector database and process documents:

   python vectorstore.py

7. Run the Django development server:

   python manage.py runserver

---

---

Usage

Interacting with the API

1. Start the server (python manage.py runserver).
2. Use tools like Postman or curl to send a POST request to the /send_message/ endpoint.
   • Endpoint: http://127.0.0.1:8000/send_message/
   • Request Body:

     {
         "question": "What are Manu's skills?"
     }

   • Response:

     {
         "response": "Manu is skilled in Python, Django, LangChain, and more..."
     }

---

API Documentation

POST /send_message/

• Description: Handles user queries and returns responses based on structured data.
• Request Body:

  {
      "question": "Your question here"
  }

• Response:
  • 200 OK: Returns the chatbot's response.
  • Example:

    {
        "response": "Manu has experience working with AI, Python, and more..."
    }

---

Directory Structure

chatbot-django-rag/
│
├── llm_utils/
│   ├── chatbot.py        # Main chatbot logic
│   └── vectorstore.py    # Vector database setup
│
├── views.py              # API endpoint logic
│
├── data/
│   ├── Resume_Data/      # Directory for resume documents
│   └── Personality_Data/ # Directory for personality documents
│
├── requirements.txt      # Python dependencies
├── manage.py             # Django management script
├── .env                  # Environment variables
└── README.md             # Project documentation

---

Contributing

1. Fork the repository.
2. Create a new branch (feature/new-feature).
3. Commit your changes.
4. Push to the branch.
5. Create a pull request.

---

License

This project is licensed under the MIT License.