AI Testing Agent Demo

A .NET-based demo that turns plain-English user stories (stored in a repo) into automated browser tests using an AI-driven two-stage strategy and Playwright. Built and demonstrated against https://www.saucedemo.com as an example target application.

This project showcases how to combine:

• Azure OpenAI (LLM) for reasoning and test-step generation
• RAG (Retrieval-Augmented Generation) for feeding repo context (user stories) to the model
• MCP (Model-Context-Protocol) as the pattern for the agent to produce actions which are executed by Playwright
• Resilient executor code patterns that correct for common AI mistakes

Purpose: a reproducible, educational demo for students and professionals to learn the foundations of autonomous test agents within the Microsoft ecosystem (Azure + .NET).

---

Presentation materials (PDF)

A full speaker-ready PDF for the talk is included in this repository:

---

Quick start for presenters

1. Prepare environment:

   • .NET 8 SDK
   • Playwright (install browsers: playwright install)
   • Azure OpenAI access (endpoint + key + deployment name)

2. Update configuration:

   • Set appsettings.json or environment variables:
     • AZURE_OPENAI_ENDPOINT, AZURE_OPENAI_DEPLOYMENT and connect to your microsoft account via visual studio
   • Ensure GitHub:UserStoryPath points to the user story you want to demo.

3. Run the demo:

   • dotnet restore && dotnet build
   • dotnet run
   • Watch the console for the two-stage planning output and Playwright browser execution.

4. Presentation flow (use the PDF; here's a short script):

   • Hook: "What if agents could do the testing for us?"
   • Demo #1: Happy path (end-to-end purchase)
   • Deep dive: Architecture (Eyes, Brain, Hands), LLM, RAG, MCP
   • Debugging story: key issues and resilient fixes
   • Q&A