AI Customer Assistant

A conversational AI support assistant for gaming support, capable of answering questions about game mechanics, player stats, clan information, and more.

Features

• Multi-agent architecture for handling different types of questions
• Combined data & knowledge retrieval for comprehensive answers
• Follow-up capabilities for ambiguous queries
• Feedback system to collect user input on answer quality
• Sample questions for quick testing
• LangSmith integration for observability and debugging

Project Structure

• Backend: FastAPI server with AI agents
• Frontend: Streamlit chat interface
• Database: SQLite with gaming support data
• Observability: LangSmith tracing and evaluation

Setup Instructions

Local Development

1. Clone the repository

   git clone https://github.com/yourusername/kgen-support-assistant.git
   cd kgen-support-assistant
   

2. Set up 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 environment variables

   • Create a .env file in the root directory

   • Add the following variables:

     OPENAI_API_KEY=your_openai_api_key
     DATABASE_URL=sqlite:///kgen_gaming_support_advanced.db
     
     # Optional LangSmith configuration for observability
     LANGCHAIN_API_KEY=your_langsmith_api_key
     LANGCHAIN_PROJECT=gaming-support-assistant
     LANGCHAIN_TRACING_V2=true
     

5. Run the backend server

   uvicorn app.main:app --reload --port 8000
   

6. Run the frontend

   cd frontend
   streamlit run streamlit_app.py
   

7. Access the application

   • Backend API: http://localhost:8000
   • Frontend: http://localhost:8501
   • LangSmith Dashboard: https://smith.langchain.com/

Deployment Options

Backend Deployment (Render)

1. Sign up for Render (https://render.com)
2. Create a new Web Service
   • Connect your GitHub repository
   • Set build command: pip install -r requirements.txt
   • Set start command: uvicorn app.main:app --host 0.0.0.0 --port $PORT
   • Add environment variables (OPENAI_API_KEY, etc.)

Frontend Deployment (Streamlit Cloud)

1. Sign up for Streamlit Cloud (https://streamlit.io/cloud)
2. Deploy your app
   • Connect your GitHub repository
   • Set the main file path: frontend/streamlit_app.py
   • Set required secrets (API_URL, DEMO_TOKEN)

Database Options

• Development: SQLite (included)
• Production: Consider migrating to PostgreSQL for better performance and scalability

API Documentation

Once the backend is running, visit:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Using the Feedback System

The application includes a feedback system that allows users to:

1. Rate responses with thumbs up/down
2. Provide detailed feedback on negative ratings
3. Track feedback in JSON files (stored in the feedback directory)

This feedback can be used to improve the AI responses over time.

Environment Configuration

Development vs Production

The application uses different settings for development and production:

• Development: Local server with debug mode
• Production: Remote servers with optimized settings

To switch between environments, set the ENVIRONMENT variable to either development or production in your .env file.

LangSmith Integration

The application integrates with LangSmith for enhanced observability, debugging, and evaluation of AI conversations. See docs/langsmith.md for detailed instructions on:

• Setting up LangSmith for your development environment
• Using traces to debug conversation flows
• Monitoring LLM performance metrics
• Evaluating and improving agent responses

LangSmith helps identify issues with:

• Query classification
• Conversation context handling
• Support ticket creation processes
• End-to-end conversation flows

License

MIT License

Credits

Built with:

• OpenAI's GPT models
• FastAPI
• Streamlit
• LangChain
• LangSmith