ruflo/examples/parallel-2

Claude-Flow Parallel Agent Testing

This example demonstrates how to run multiple Claude-Flow SPARC agents in parallel to test their capabilities and measure performance.

Overview

The parallel test coordinator runs 8 different agent types simultaneously:

• Specification Agent: Creates detailed API specifications
• Architecture Agent: Designs system architectures
• Code Agent: Implements algorithms and data structures
• TDD Agent: Follows Test-Driven Development practices
• Debug Agent: Analyzes and fixes code issues
• Documentation Agent: Creates comprehensive documentation
• Security Review Agent: Analyzes security vulnerabilities
• Integration Agent: Plans system integration strategies

Installation

# Navigate to the parallel-2 directory
cd examples/parallel-2

# Install dependencies
npm install

Usage

Run the Parallel Test

npm run test

This will:

1. Launch all agents in parallel
2. Execute their respective tasks
3. Save results to results.json

Analyze Results

npm run analyze

This will:

1. Read the test results
2. Generate performance statistics
3. Create a detailed report in detailed-report.md

Full Test Cycle

npm run test:full

This runs the complete test cycle: clean, test, and analyze.

Test Configuration

Each agent is given a specific task designed to test its capabilities:

Agent	Task Description
Specification	Create calculator API specification
Architecture	Design microservices for e-commerce
Code	Implement binary search tree
TDD	Create string utility library with tests
Debug	Analyze and fix memory leaks
Documentation	Create API documentation
Security	Review authentication implementation
Integration	Plan microservice integration

Understanding Results

results.json

Contains raw test execution data:

• Agent execution times
• Success/failure status
• Output or error messages
• Overall statistics

detailed-report.md

Provides comprehensive analysis:

• Executive summary
• Performance metrics
• Execution timeline
• Efficiency calculations
• Recommendations

Performance Metrics

The test suite measures:

• Individual agent duration: Time taken by each agent
• Parallel efficiency: Comparison with sequential execution
• Success rate: Percentage of successful agent executions
• Resource utilization: Overall system performance

Extending the Tests

To add new test cases:

1. Create a new test file in test-agents/:

export const newTestCase = {
  name: "Test Name",
  description: "Test description",
  task: "Detailed task for the agent",
  expectedOutputs: ["Expected output 1", "Expected output 2"],
  validateOutput: (output: string): boolean => {
    // Validation logic
    return true;
  }
};

2. Add the test to parallel-test.ts:

{
  name: "New Agent",
  mode: "agent-mode",
  task: "Your task description",
  priority: 1
}

Troubleshooting

Common Issues

1. Agent timeout: Increase timeout in coordinator.ts (default: 5 minutes)
2. Missing results: Ensure all agents complete before analyzing
3. Permission errors: Check write permissions for output files

Debug Mode

To see detailed agent output, modify coordinator.ts:

console.log(`Agent ${task.name} output:`, stdout);

Architecture

parallel-2/
├── coordinator.ts       # Main coordination logic
├── parallel-test.ts     # Test configuration and runner
├── analyze-results.ts   # Results analysis tool
├── test-agents/        # Individual agent test cases
│   ├── spec-test.ts
│   ├── architect-test.ts
│   ├── code-test.ts
│   ├── tdd-test.ts
│   └── debug-test.ts
├── results.json        # Test execution results
└── detailed-report.md  # Analysis report

Best Practices

1. Resource Management: Monitor system resources during parallel execution
2. Task Complexity: Balance task complexity across agents
3. Error Handling: Implement proper error recovery
4. Logging: Enable detailed logging for debugging
5. Scaling: Consider batching for large numbers of agents

Contributing

To contribute new test cases or improvements:

1. Follow the existing test structure
2. Ensure tests are meaningful and measurable
3. Update documentation accordingly
4. Test thoroughly before submitting

License

This example is part of the Claude-Flow project and follows the same license terms.