Predator: The Autonomous Sales Orchestrator 🦅

📖 Overview

Predator is an enterprise-grade AI agent built with the Google ADK. It automates the "research-to-outreach" sales loop by using a multi-agent architecture to score leads, check competitor pricing (via A2A), and draft emails that require human approval (Human-in-the-Loop).

🏗️ Architecture

• Brain: Gemini 2.5 Flash-Lite
• Framework: Google Agent Development Kit (ADK)
• Protocol: Agent-to-Agent (A2A) for market data
• Deployment: Vertex AI Agent Engine / Cloud Run

🚀 Key Features (Course Concepts Applied)

1. Multi-Agent System: Orchestrator delegates to a 'Market Oracle' agent.
2. A2A Communication: Uses RemoteA2aAgent to query external services.
3. Safety First: Implements tool_context.request_confirmation for email tools.
4. Memory Management: Uses EventsCompactionConfig to summarize long context.

🛠️ How to Run

1. Clone the repo.
2. Install dependencies: pip install -r requirements.txt
3. Start the Market Oracle (A2A Service):

   python market_oracle.py

4. Run Predator:

   python predator.py

📊 Deployment

This agent is configured for deployment on Google Cloud Vertex AI Agent Engine. See .agent_engine_config.json for resource limits.