OpenAgentsControl/CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

0.1.0-alpha.1 - 2025-11-26

[0.3.0] - 2025-11-26

Changes

• feat(ci): skip redundant tests on PR merge

Tests only run on:

• Pull requests (CI checks)
• Direct pushes to main
• Manual workflow dispatch

PR merges skip tests (already passed) but still trigger version bump.

[0.2.0] - 2025-11-26

Changes

• feat(ci): auto-create GitHub releases on version bump

Releases now appear in the GitHub sidebar. The release notes are extracted from CHANGELOG.md for the specific version.

[0.1.0] - 2025-11-26

Changes

• fix(ci): add contents:write permission for auto version bump

The GITHUB_TOKEN needs explicit write permission to push commits and tags back to the repository.

Added

SDK-Based Evaluation Framework

• Complete test execution framework using OpenCode SDK
• Support for openagent and opencoder testing
• Real agent testing with session management
• Smart timeout system with activity monitoring
• Multi-turn conversation support

Modular Architecture

• Refactored test-runner.ts (884 lines → 4 focused modules):
  • test-runner.ts (411 lines): Thin orchestrator
  • test-executor.ts (392 lines): Core execution logic
  • result-validator.ts (253 lines): Validation logic
  • event-logger.ts (128 lines): Logging utilities
• Improved Single Responsibility Principle compliance
• Enhanced testability through dependency injection

Test Infrastructure

• 20+ test cases across multiple categories:
  • OpenAgent: Developer (12), Context Loading (5), Business (2), Edge Cases (3)
  • OpenCoder: Developer (4)
• BehaviorEvaluator for validating expected agent actions
• Comprehensive evaluators: approval-gate, context-loading, delegation, tool-usage

Interactive Results Dashboard

• Real-time test results visualization
• Filtering by agent, category, status
• Detailed violation tracking
• CSV export functionality
• Historical results tracking
• One-command deployment (./serve.sh)

Documentation

• ARCHITECTURE.md: Comprehensive system review (456 lines)
• GETTING_STARTED.md: Quick start guide (435 lines)
• SDK_EVAL_README.md: Complete SDK guide (298 lines)
• Test design guide and architecture overview
• Documentation cleanup (removed 3 outdated files)

Script Organization

• Organized 12 scripts into logical directories:
  • scripts/debug/: Session debugging tools (4 files)
  • scripts/test/: Test execution scripts (6 files)
  • scripts/utils/: Utility scripts (2 files)
• Comprehensive scripts/README.md with usage examples

Monorepo Structure

• Root package.json with convenient npm scripts
• Easy agent selection (openagent, opencoder)
• Easy model selection (grok, claude, gpt-4)
• Quick dashboard access from root
• No folder navigation required

CI/CD

• GitHub Actions workflow for automated testing
• Pre-merge validation for agent changes
• Fast smoke tests for both agents
• Automated test result reporting

Agent Improvements

• Enhanced openagent with better context loading
• New opencoder agent with test suite
• Improved subagent invocation patterns
• Ultra-compact context index system

Changed

• Reorganized evaluation framework structure
• Improved test case schema with behavior expectations
• Enhanced context loading detection

Removed

• Outdated documentation files (TESTING_CONFIDENCE.md, TEST_REVIEW.md, SESSION_STORAGE_FIX.md)
• Redundant test files

Fixed

• Context loading evaluator detection accuracy
• Multi-turn prompt handling
• Test artifact cleanup

---

Version Format

v0.1.0-alpha.1
│ │ │  │      │
│ │ │  │      └─ Build/Iteration number
│ │ │  └──────── Release stage (alpha, beta, rc)
│ │ └─────────── Patch version
│ └───────────── Minor version
└─────────────── Major version (0 = pre-release)

Version Progression

• Alpha (v0.x.0-alpha.N): Early development, unstable
• Beta (v0.x.0-beta.N): Feature complete, testing
• RC (v0.x.0-rc.N): Release candidate, stable
• Stable (v1.x.x): Production ready