Skip to content

README.md

Latest commit

 

History

History
51 lines (34 loc) · 1.79 KB

README.md

File metadata and controls

51 lines (34 loc) · 1.79 KB

Tutorial Tester Agent

An AI-powered agent that automatically tests and validates tutorials using browser automation and intelligent decision-making.

🚀 Features

  • Browser Automation: Uses Playwright MCP Server for realistic browser interactions
  • AI-Powered Testing: LangGraph ReAct agent with Claude 4 Sonnet through SAP's Generative AI Hub
  • Tutorial Validation: Automatically executes tutorial steps and identifies issues

🛠️ Technology Stack

  • LLM: Claude 4 Sonnet access through SAP's Generative AI Hub (Cloud SDK for AI - Python)
  • Agent Framework: LangGraph with ReAct pattern
  • Browser Automation: Playwright MCP Server

📁 Project Structure

tutorial-system-agent/
├── tutorial-ai-agent/
│   ├── agent/
│   │   └── tutorial_tester_agent.py    # Main agent
│   └── data/
│       ├── tutorials/                  # Tutorial markdown files
│       └── output/                     # JSON results
├── requirements.txt                    # Python dependencies
├── docker-compose.yml                  # Container orchestration
└── Dockerfile                         # Container definition

🎯 How It Works

  1. Load Tutorial: Reads markdown tutorial files
  2. Initialize Agent: Creates LangGraph ReAct agent with MCP tools
  3. Execute Steps: Agent follows tutorial instructions in browser
  4. Validate Results: Checks for completion and identifies issues
  5. Generate Report: Saves structured JSON logs with timestamps

🔧 Configuration

Key settings in tutorial_tester_agent.py:

  • TUTORIAL_FILE: Tutorial to test (default: "ailaunchpad-orchestration.md")
  • RECURSION_LIMIT: Max agent steps (default: 500)
  • MODEL_ID: LLM model (default: "anthropic--claude-4-sonnet")