Cotor logoCotor
docs/UPGRADE_GUIDE.md

Cotor Upgrade Guide

open_in_new원문

Cotor Upgrade Guide

What's New in Cotor 1.0

🎯 Major Improvements

1. Enhanced CLI Commands

Before:

cotor board-implementation  # Confusing, inconsistent

After:

cotor run board-implementation              # Clear and intuitive
cotor run board-implementation --verbose    # Detailed logging
cotor run board-implementation --dry-run    # Simulation mode
cotor run board-implementation --watch      # Real-time monitoring

2. Real-Time Progress Monitoring

The new pipeline monitor shows live execution progress:

🚀 Running: board-implementation (5 stages)
┌──────────────────────────────────────────────┐
│ ✅ Stage 1: requirements-analysis      2.3s  │
│ 🔄 Stage 2: backend-implementation           │
│ ⏳ Stage 3: code-review                      │
│ ⏳ Stage 4: testing                          │
│ ⏳ Stage 5: documentation                    │
└──────────────────────────────────────────────┘
⏱️  Elapsed: 00:02:34 | Progress: 20% (1/5 stages completed)

3. Pipeline Validation

Validate your pipeline before running:

cotor validate board-implementation

# Output:
✅ Pipeline structure: valid
✅ All agents defined: valid
✅ Stage dependencies: valid
⚠️  Warning: Stage 'backend-implementation' has no timeout specified

4. User-Friendly Error Messages

Before:

Error: Failed to load config

After:

❌ Error: Pipeline configuration not found

📍 Problem:
   cotor.yaml file is missing in the current directory

💡 Solutions:
   1. Run 'cotor init' to create a default configuration
   2. Specify config path: cotor run -c path/to/config.yaml <pipeline>
   3. Check if you're in the correct directory

📖 Documentation: https://docs.cotor.dev/configuration

5. Dry-Run Mode

Test your pipeline without actually executing it:

cotor run board-implementation --dry-run

# Output:
📋 Pipeline Estimate: board-implementation
   Execution Mode: SEQUENTIAL

Stages:
  ├─ requirements-analysis (claude)
  │  └─ ~30s
  ├─ backend-implementation (claude)
  │  └─ ~45s
  ├─ code-review (gemini)
  │  └─ ~30s
  ├─ testing (gemini)
  │  └─ ~40s
  ├─ documentation (claude)
  │  └─ ~25s

⏱️  Total Estimated Duration: ~2m 50s

🛠️ Technical Improvements

New Dependencies

Added for better terminal UI and progress tracking:

implementation("com.github.ajalt.mordant:mordant:2.2.0")  // Terminal UI
implementation("me.tongfei:progressbar:0.9.5")            // Progress bar
implementation("io.github.microutils:kotlin-logging-jvm:3.0.5") // Logging

New Components

  1. PipelineMonitor (com.cotor.monitoring.PipelineMonitor)

    • Real-time stage tracking
    • Progress visualization
    • Duration calculation
    • Colored terminal output

  2. UserFriendlyError (com.cotor.error.UserFriendlyError)

    • Structured error messages
    • Solution suggestions
    • Documentation links

  3. PipelineValidator (com.cotor.validation.PipelineValidator)

    • Configuration validation
    • Dependency checking
    • Duration estimation
    • Critical path analysis

  4. Enhanced Commands (com.cotor.presentation.cli.EnhancedCommands)

    • EnhancedRunCommand - Improved pipeline execution
    • ValidateCommand - Pipeline validation
    • TestCommand - Testing framework

🔁 Conditional & Loop Pipelines

  • Stage types – PipelineStage.type defaults to EXECUTION, but can now be DECISION or LOOP.
  • Decision stages – condition.expression uses the new evaluator to read previous stage metadata/output (review.validationScore >= 0.9) and pick an action (CONTINUE, GOTO, ABORT). sharedState assignments let you pass values to later prompts.
  • Loop stages – loop config targets any stage, enforces maxIterations, and optionally breaks once untilExpression becomes true.
  - id: review
    agent:
      name: gemini

  - id: gate
    type: DECISION
    condition:
      expression: "review.validationScore >= 0.85"
      onTrue:
        action: CONTINUE
      onFalse:
        action: GOTO
        targetStageId: improve

  - id: loop-controller
    type: LOOP
    loop:
      targetStageId: review
      maxIterations: 3
      untilExpression: "improve.validationScore >= 0.92"

ℹ️ Conditional stages are supported in SEQUENTIAL pipelines. Parallel/DAG pipelines continue to run pure execution stages.

🧠 Result Intelligence

  • ResultAnalyzer – new component (com.cotor.analysis.ResultAnalyzer) compares agent outputs and metadata.
  • Run Summary Enhancements – CLI now prints a consensus badge, confidence percentage, and highlights the best agent.
  • Diagnostics – verbose mode reveals which agents disagree and provides remediation tips (e.g., rerun strongest agent).

Example summary:

📦 Run Summary
   Pipeline : benchmark
   Agents   : 4/4 succeeded
   Duration : 11.4s
   Consensus: ⚠️ Divergent (48%)
   Best     : gemini - Added pagination defaults plus schema notes.

Event System Updates

New events for stage-level monitoring:

data class StageStartedEvent(val stageId: String, val pipelineId: String)
data class StageCompletedEvent(val stageId: String, val pipelineId: String, val result: AgentResult)
data class StageFailedEvent(val stageId: String, val pipelineId: String, val error: Throwable)

🚀 Migration Guide

Step 1: Update Your Scripts

Replace old command patterns:

# Old
./cotor board-implementation

# New
./cotor run board-implementation

Step 2: Add Validation

Add validation step to your CI/CD:

# Validate before running
./cotor validate my-pipeline
./cotor run my-pipeline

Step 3: Use Verbose Mode for Debugging

When debugging issues:

./cotor run my-pipeline --verbose

Step 4: Test with Dry-Run

Before running expensive pipelines:

./cotor run my-pipeline --dry-run

📊 Performance Improvements

  • Parallel Event Emission: Events are now emitted asynchronously
  • Efficient Progress Rendering: Terminal updates only when state changes
  • Smart Caching: Agent registry caches loaded agents
  • Memory Management: Proper cleanup after pipeline execution

🔒 Security Enhancements

  • Input Validation: All pipeline configurations are validated
  • Timeout Enforcement: Stages respect configured timeouts
  • Error Isolation: Stage failures don't crash the entire system
  • Safe Execution: Dry-run mode for testing without side effects

📖 New Documentation

  • UPGRADE_GUIDE.md (this file) - Upgrade instructions
  • CLI_REFERENCE.md - Complete CLI command reference
  • VALIDATION_GUIDE.md - Pipeline validation guide
  • ERROR_HANDLING.md - Error handling best practices

🧪 Testing

New test suite for validation:

./test-cotor-enhanced.sh

# Test specific features
./cotor test --test-dir test/board-feature

🎯 Roadmap

Phase 2 (Coming Soon)

  • Result artifact management
  • Pipeline resume from failure point
  • Web dashboard enhancements
  • Performance profiling

Phase 3 (Future)

  • Parallel execution optimization
  • Distributed pipeline execution
  • Advanced caching strategies
  • Plugin marketplace

📞 Support

🙏 Acknowledgments

This upgrade incorporates feedback from the community and implements industry best practices from tools like:

  • Gradle - Progress visualization
  • Terraform - Plan/Apply separation
  • Kubernetes - Error messaging