PBI Manual Test Case Generator

A web application for generating manual test cases for Product Backlog Items (PBIs) from Azure DevOps using Claude AI.

Quick Start

1. First Time Setup

# Install dependencies
npm install

# Login to Azure CLI
az login

That's it! The app uses AWS Bedrock (already configured) for Claude AI.

2. Launch

Double-click: Launch PBI-PR Analyzer.bat

The application will automatically:

• Build the Angular frontend (first run only)
• Start the server on port 3000
• Open your browser

That's it! Start generating test cases for PBIs.

Features

• Fetch PBIs from Azure DevOps automatically
• AI-Powered Manual Test Case Generation using Claude Opus 4.6:
  • Generate comprehensive manual test cases following GL Assessment standards
  • Context-aware using embedded knowledge base
  • Includes functional, UAT, accessibility, and security test scenarios
  • Step-by-step instructions ready for manual testers
  • Happy Path first - Test Case #1 is always the successful scenario
• Iterative Quality Improvement:
  • Claude self-reviews generated test cases against your Definition of Done
  • Automatically iterates (up to 3 times) to improve quality
  • Provides quality score (1-10) with justification
  • Shows how many iterations were needed
• User Feedback Loop:
  • Provide specific feedback if test cases need improvement
  • Regenerate with your feedback incorporated
  • Iterative refinement until you're satisfied
• Embedded Knowledge Base - No external dependencies required:
  • GL Assessment testing standards (DoR/DoD)
  • Platform architecture and tech stack
  • Domain glossary (CAT4, NGRT, SAS, etc.)
  • Example PBIs and test cases (optional)
  • Definition of Done criteria (optional)
• Azure CLI Authentication - Uses your existing az login session
• Simple Web App - Runs in your browser, no desktop app issues

How to Use

1. Enter PBI ID (e.g., 95962) and Project (e.g., Testwise)
2. Click Fetch PBI
3. Review the PBI details displayed
4. Click Generate Manual Test Cases
5. Claude generates and self-reviews (with iterative improvement if Definition of Done is provided):
   • Generates initial test cases
   • Reviews against quality criteria
   • Improves and refines (up to 3 iterations)
   • Provides quality score and feedback
6. Review the results:
   • View quality score (if Definition of Done is provided)
   • See how many iterations were needed
   • Read comprehensive manual test cases
7. Optional: Provide feedback and regenerate:
   • Enter specific feedback in the textarea
   • Click "Regenerate with Feedback"
   • Claude incorporates your feedback and generates improved version
8. Copy final test cases to clipboard for your test plan

Architecture

Browser (Angular) → Express Server → Azure DevOps API
                                   → Claude API

Simple web application with:

• Frontend: Angular running in your browser
• Backend: Express.js server handling API calls
• Authentication: Azure CLI for DevOps, environment variable for Claude

Prerequisites

• Node.js (v18+)
• npm (v9+)
• Azure CLI (az) with active login
• AWS credentials configured (for Bedrock access to Claude)

Configuration

The application automatically uses:

• Azure CLI token from az login for Azure DevOps access
• AWS Bedrock with your configured AWS credentials for Claude AI access
  • Region: us-east-1
  • Profile: ai-tools-prod
  • Model: us.anthropic.claude-opus-4-6-v1

View current status in the Settings page:

• Azure DevOps Organization (default: https://dev.azure.com/gl-development)
• Default Project (e.g., Testwise)
• AWS Bedrock configuration
• Authentication status

Embedded Knowledge Base

This application includes a curated knowledge base in the knowledge/ folder that enhances test case generation:

What's Included

• testing-standards.md - GL Assessment testing standards:

  • Definition of Ready (DoR) and Definition of Done (DoD)
  • Acceptance criteria format (Given/When/Then)
  • Required test types (Cypress, SpecFlow, Jest, TestContainers)
  • Browser compatibility requirements

• glossary.md - Domain terminology:

  • Product acronyms (CAT4, NGRT, PASS, etc.)
  • Technical terms (SAS, NPR, Stanine, Batman, Epoch, etc.)
  • Infrastructure and repository names

• platform-overview.md - Technical context:

  • Platform architecture and components
  • Technology stack (Angular, .NET, Azure)
  • Authentication patterns
  • Integration points

• examples/ - Example PBIs and test cases (Optional):

  • Place example PBI JSON files in knowledge/examples/pbis/
  • Place corresponding test case markdown files in knowledge/examples/test-cases/
  • Files must have matching names (e.g., login.json and login.md)
  • Claude learns from these examples to match your quality and format
  • See examples/README.md for instructions

• definition-of-done/ - Quality criteria (Optional):

  • Place Definition of Done markdown files in this folder
  • Defines what makes test cases "ready for use"
  • Claude uses these for self-review and iterative improvement
  • Enables quality scoring (1-10) of generated test cases
  • See definition-of-done/README.md for instructions

How It Works

When you generate test cases:

1. Context Loading: Claude receives the knowledge base, examples (if any), and Definition of Done (if any)

2. Initial Generation: Claude creates test cases using all available context

3. Self-Review (if Definition of Done exists):

   • Claude reviews its own output against quality criteria
   • Identifies issues and areas for improvement

4. Iterative Improvement (up to 3 iterations):

   • Claude generates improved version addressing issues
   • Process repeats until quality standards are met or max iterations reached

5. Quality Scoring (if Definition of Done exists):

   • Claude provides quality score (1-10) with justification
   • Lists strengths and areas for improvement

6. User Feedback (optional):

   • You can provide specific feedback
   • Claude regenerates incorporating your feedback
   • Process can repeat as needed

No external dependencies required - everything is self-contained in this repository.

Troubleshooting

Application won't start

• Check Node.js is installed: node --version
• Ensure dependencies are installed: npm install

"Azure CLI not logged in"

az login

Claude analysis fails

Check that your AWS credentials are configured and you have access to Bedrock:

aws sts get-caller-identity --profile ai-tools-prod

Port 3000 already in use

Change the PORT variable in server.js (line 9)

Build fails

npm install
npm run build

Project Structure

pbi-test-case-generator/
├── Launch PBI-PR Analyzer.bat    # Launcher
├── server.js                     # Express backend
├── package.json                  # Dependencies
├── src/app/
│   ├── components/
│   │   ├── analysis/             # Analysis UI
│   │   └── settings/             # Settings page
│   └── services/
│       └── api.service.ts        # HTTP API client
└── dist/                         # Built Angular app

Scripts

npm run start:web    # Start the server
npm run build        # Build Angular app
npm run build:prod   # Production build
npm run start        # Angular dev server (for development)

Technologies

• Angular 20 - Frontend framework
• Express.js - Backend server
• Azure DevOps Node API - DevOps integration
• Anthropic Claude SDK - AI analysis

Benefits

✅ No antivirus conflicts
✅ Faster startup (~1-2 seconds)
✅ Smaller size (~230MB vs ~550MB for desktop apps)
✅ Better debugging with browser DevTools
✅ Works on any machine with Node.js

Support

For issues or questions, contact the development team.

---

Ready to start? → Double-click Launch PBI-PR Analyzer.bat