Thoughtful AI Customer Support Agent

A simple conversational AI agent that answers questions about Thoughtful AI's healthcare automation products.

🎬 Demo

Watch the application in action:

https://github.com/mrjoema/ai-agent-chat-challenge/blob/main/demo.mov

Demo video showing the chat interface, question matching, fallback responses, and error handling in action.

Note: If the video doesn't play above, you can download and view demo.mov locally from the repository.

Features

• Intelligent Question Matching: Uses TF-IDF vectorization for finding relevant answers
• LLM Fallback: Uses OpenAI API for questions not in the knowledge base
• Robust Error Handling: Gracefully handles API failures, timeouts, and invalid inputs
• Security: Input validation and sanitization to prevent harmful inputs
• User-Friendly Interface: Clean Streamlit chat interface

Quick Start

🚀 Easy Setup (Recommended)

1. Run the startup script:

   ./run.sh

   This script will automatically:

   • Create a virtual environment if needed
   • Install all dependencies
   • Copy .env template if needed
   • Start the Streamlit application

2. Open your browser to http://localhost:8501

3. (Optional) Add OpenAI API key for enhanced responses:

   • Edit the .env file and add your OpenAI API key
   • Restart the application

🔧 Manual Setup

If you prefer manual setup:

1. Create virtual environment and install dependencies:

   python3 -m venv venv
   source venv/bin/activate
   python3 -m pip install -r requirements.txt

2. Set up environment:

   cp .env.example .env
   # Edit .env and add your OpenAI API key (optional)

3. Run the application:

   streamlit run app.py

Usage

• Ask questions about Thoughtful AI's products (EVA, CAM, PHIL)
• The agent will provide relevant answers or use AI-generated responses for other questions
• Works without OpenAI API key using hardcoded responses and keyword matching

Example Questions

• "What does EVA do?"
• "Tell me about claims processing"
• "How does PHIL work?"
• "What are the benefits of using Thoughtful AI's agents?"

Architecture

Core Components

• agent.py: Core logic including input validation, question matching, and LLM integration
• app.py: Streamlit user interface with chat functionality
• Multi-layer fallback: TF-IDF → Keyword matching → LLM → Generic response

Error Handling

The agent handles:

• Invalid or harmful inputs (XSS, SQL injection attempts)
• API timeouts and rate limits
• Connection failures
• Missing API keys
• Unexpected errors

All errors are logged and users receive helpful fallback messages.

Security Features

• Input length validation
• Harmful pattern detection
• Content sanitization
• Rate limiting
• Response validation

Configuration

Environment variables (optional):

• OPENAI_API_KEY: Your OpenAI API key for enhanced responses
• LOG_LEVEL: Logging level (default: INFO)
• CONFIDENCE_THRESHOLD: Matching confidence threshold (default: 0.3)
• MAX_INPUT_LENGTH: Maximum input length (default: 500)

Knowledge Base

The agent currently knows about:

• EVA: Eligibility Verification Agent
• CAM: Claims Processing Agent
• PHIL: Payment Posting Agent
• General benefits of Thoughtful AI's automation solutions

Technical Details

• Framework: Streamlit for web UI
• NLP: scikit-learn TF-IDF vectorization
• Matching: Cosine similarity with fallback to keyword matching
• LLM: OpenAI GPT-3.5-turbo for general questions
• Error Handling: Comprehensive retry logic and graceful degradation

Development

The codebase demonstrates:

• Clean, modular architecture
• Comprehensive error handling
• Security best practices
• Professional UI/UX
• Easy deployment and configuration

Perfect for showcasing software engineering skills in a time-constrained environment!