agentic-ai-mcp-workflows

Insurance AI POC - Google ADK v1.2.1

Production-ready insurance AI agent system built with Google's official Agent Development Kit

🚀 Overview

This is a complete production implementation of an insurance AI agent system using Google's official Agent Development Kit (ADK) v1.2.1. The system follows patterns from google/adk-samples and eliminates the need for custom FastAPI servers.

✨ Key Features

• 🤖 Multiple Agent Types: Customer service (LlmAgent) + Technical operations (BaseAgent)
• 🛠️ Native Google ADK: Uses built-in adk web, adk run, and adk api_server commands
• 🎛️ No Custom FastAPI: Leverages Google ADK's built-in runtime and web UI
• 📊 Built-in Evaluation: Google ADK evaluation framework included
• 🚀 Production Ready: Deployment tools and monitoring integrated
• 📝 Automatic Logging: Built-in tracing and observability

🏗️ Architecture

insurance-adk/
├── insurance_customer_service/    # LlmAgent for customer interactions
│   ├── __init__.py
│   └── agent.py                  # root_agent = LlmAgent(...)
├── insurance_technical_agent/     # BaseAgent for backend operations  
│   ├── __init__.py
│   └── agent.py                  # root_agent = BaseAgent(...)
├── agent.py                      # Simple agent example
└── .env                          # Google API configuration

🛠️ Google ADK Commands

Available Commands

# Interactive CLI mode
adk run insurance_customer_service
adk run insurance_technical_agent

# Web UI for testing (recommended)
adk web

# REST API server  
adk api_server

# Evaluation framework
adk eval <agent> <eval_set>

# Create new agents
adk create <app_name>

🚦 Quick Start

Prerequisites

# Install dependencies
pip install -r requirements.txt

# Set up Google API key
export GOOGLE_API_KEY=your_google_ai_studio_api_key
# OR configure Vertex AI in .env file

Run Agents

cd insurance-adk

# Option 1: Web UI (recommended)
adk web

# Option 2: CLI mode
adk run insurance_customer_service

# Option 3: API server
adk api_server

🎯 Agent Capabilities

Insurance Customer Service Agent (LlmAgent)

• 💬 Customer support and inquiries
• 📋 Policy information assistance
• 🤝 Claims guidance and support
• 📞 Professional customer interactions

Insurance Technical Agent (BaseAgent)

• ⚙️ Complex backend operations
• 📊 Policy analysis and validation
• 🔧 Technical system integrations
• 📈 Data processing workflows

📊 Benefits Over FastAPI

Google ADK	Custom FastAPI
✅ Built-in web UI	❌ Manual UI development
✅ Automatic evaluation	❌ Custom testing framework
✅ Production deployment	❌ Manual deployment setup
✅ Integrated logging	❌ Custom logging configuration
✅ CLI tools included	❌ Manual CLI development
✅ Official Google support	❌ Community maintenance

🔧 Configuration

Environment Variables (.env)

# Google AI Configuration
GOOGLE_API_KEY=your_google_ai_studio_api_key

# OR Vertex AI (production)
GOOGLE_GENAI_USE_VERTEXAI=TRUE
GOOGLE_CLOUD_PROJECT=your-project-id
GOOGLE_CLOUD_LOCATION=us-central1

# Model Configuration  
MODEL_NAME=gemini-2.0-flash

🧪 Testing

# Run Google ADK tests
python tests/google-adk-tests/test_google_adk_agents.py

# Demo commands
python tests/google-adk-tests/demo_adk_commands.py

# A2A communication tests  
python tests/google-adk-tests/test_a2a_communication.py

📈 Monitoring & Observability

• Langfuse: LLM observability and tracing
• Google ADK Logs: Built-in agent logging
• Evaluation Framework: Automated testing suite

🌟 Production Features

• 🔄 Hot Reloading: Automatic agent updates
• 📊 Built-in Metrics: Performance monitoring
• 🚀 Easy Deployment: adk deploy commands
• 🔒 Security: Google Cloud integration
• 📝 Documentation: Auto-generated API docs

🎉 Migration Complete

✅ FastAPI Server Removed: No longer needed - Google ADK provides all runtime capabilities
✅ Agent Structure: Following google/adk-samples patterns
✅ Built-in Tools: Using native adk commands instead of custom scripts
✅ Clean Architecture: Simplified and production-ready

---

Ready to use Google ADK v1.2.1! 🚀

Start with: cd insurance-adk && adk web