LangGraph Chatbot with Chainlit Interface

This project implements a conversational AI chatbot using LangGraph for conversation management and Chainlit for the web interface. The chatbot features automatic conversation summarization to maintain context while keeping memory usage efficient.

Features

• LangGraph Workflow: Uses LangGraph for managing conversation state and flow
• Automatic Summarization: Automatically summarizes conversations when they exceed 6 messages
• Memory Management: Efficiently manages conversation history using checkpoints
• Modern Web Interface: Beautiful Chainlit-based chat interface
• Ollama Integration: Uses local Ollama models (phi3:mini by default)

Architecture

The system consists of two main components:

1. app.py: Contains the core LangGraph workflow logic

   • Conversation state management
   • Automatic summarization
   • Message processing pipeline

2. chainlit.py: Provides the web interface

   • User-friendly chat interface
   • Session management
   • Integration with the LangGraph workflow

Prerequisites

• Python 3.8+
• Ollama installed and running locally
• phi3:mini model downloaded (ollama pull phi3:mini)

Installation

1. Clone the repository and navigate to the project directory:

cd chatbot_with_summarization

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. Ensure Ollama is running and the model is available:

ollama serve
ollama pull phi3:mini

Usage

Running the Chainlit Interface

Start the Chainlit chatbot:

chainlit run chainlit.py

The chatbot will be available at http://localhost:8000

Testing the Core Logic

You can also test the core LangGraph workflow directly:

python app.py

How It Works

1. Conversation Start: When a user starts chatting, a unique thread ID is created
2. Message Processing: Each message is processed through the LangGraph workflow
3. Automatic Summarization: When the conversation exceeds 6 messages, the system automatically summarizes the conversation
4. Memory Cleanup: Old messages are removed, keeping only the most recent ones
5. Context Preservation: The summary maintains conversation context for future interactions

Configuration

Model Settings

You can modify the model settings in app.py:

def create_model():
    return ChatOllama(
        model="phi3:mini",  # Change model here
        temperature=0.8,     # Adjust creativity
        num_predict=256,     # Adjust response length
    )

Summarization Threshold

Adjust when summarization occurs by modifying the should_continue function:

def should_continue(state: State) -> Literal["summarize_conversation", END]:
    messages = state["messages"]
    if len(messages) > 6:  # Change this number
        return "summarize_conversation"
    return END

Chainlit UI

Customize the interface appearance in .chainlit/config.toml:

• Theme (light/dark)
• Assistant name and description
• Chat features and settings

Project Structure

chatbot_with_summarization/
├── app.py                 # Core LangGraph workflow
├── chainlit.py           # Chainlit web interface
├── requirements.txt      # Python dependencies
├── .chainlit/           # Chainlit configuration
│   └── config.toml
└── README.md            # This file

Troubleshooting

Common Issues

1. Ollama Connection Error: Ensure Ollama is running (ollama serve)
2. Model Not Found: Download the required model (ollama pull phi3:mini)
3. Port Already in Use: Chainlit uses port 8000 by default. Change it with chainlit run chainlit.py --port 8001

Performance Tips

• Use smaller models for faster responses
• Adjust the summarization threshold based on your needs
• Monitor memory usage with very long conversations

Contributing

Feel free to submit issues and enhancement requests!

License

This project is open source and available under the MIT License.