Releases: apiad/opencode
Releases · apiad/opencode
Release v1.0.0
What's Changed
Added
- Build Agent with Full TCR Workflow: Implemented a complete Test-Commit-Revert agent for test-driven implementation with single-iteration workflow.
- Task Command: Added dedicated
/taskcommand for managing project roadmap intasks.yaml. - Plan Workflow Improvements: Enhanced task linking workflow with better plan synchronization.
Changed
- Maintenance Command: Reassigned from
queryagent toplanagent for improved audit workflows. - Scaffold Command: Streamlined by removing execution phase from command spec.
Refactored
- TypeScript/YAML Migration: Migrated task and journal tools from Python to TypeScript with YAML storage (BREAKING).
- Build Agent Simplification: Simplified to one-off TCR agent, removing preconditions and constraints.
- Subagent Structure: Flattened subagent directory structure in
.opencode/.
Documentation
- Updated all references from Python tools to TypeScript/YAML.
- Updated documentation to reference
.opencodepaths. - Added context minification feature documentation and principles.
Full Changelog: v0.22.0...v1.0.0
Release v0.22.0
Release v0.22.0
v0.20.0
[0.20.0] - 2026-03-23
Added
- Unified Discovery-Plan-Execute Lifecycle: Formally established a strict three-phase architectural boundary (Discovery -> Plan -> Execution).
- Modernized Test Suite (uv/pytest): Replaced legacy testing infrastructure with
uvfor dependency management andpytestfor automated validation. - Dynamic Structural Testing: Implemented
tests/test_structure.pyfor discover-based, schema-compliant verification of all 15 commands and 6 agents. - Documentation Overhaul: Updated the entire documentation suite (
docs/index.md,docs/design.md,docs/develop.md,docs/user-guide.md) to reflect the new unified lifecycle and technical stack. - Improved Content Creation: Refactored the
/draftcommand and evolved thewriteragent for review-driven refinement.
Changed
- Read-Only Audit Workflows: Refactored
/maintenanceand/review(formerly/revise) to be strictly read-only 'Discovery' commands that produce artifacts for/plan. - Review Command Evolution: Evolved the
/reviewcommand into a non-destructive, multi-phase audit workflow with sidecar reporting. - Agent Naming Alignment: Renamed
editortoreviewerandreportertowriterfor improved semantic clarity. - Test Optimization: Refactored
tests/test_review_command.pyandtests/test_sync.pyto be dynamic and less brittle, removing hardcoded skips and string matches.
Release v0.19.1
Changed
- Documentation Suite Overhaul: Major updates to the entire documentation suite (
index.md,deploy.md,design.md,develop.md,user-guide.md, andREADME.md) to accurately reflect the framework's v0.19.0 improvements, including the/learncommand, the dual-layer hook system, and the updated testing philosophy.
v0.18.0
[0.18.0] - 2026-03-20
Added
- Installer Boilerplate Generation:
docs/install.shnow generates clean, generic boilerplate forREADME.md,CHANGELOG.md,TASKS.md, andmakefileinstead of copying project-specific versions when scaffolding a new project. - Scientific Debugging Documentation: Added comprehensive documentation for the
/debugcommand's scientific, hypothesis-driven workflow indocs/design.mdanddocs/user-guide.md.
Changed
- Documentation Suite Overhaul: Major updates to the entire documentation suite (
index.md,deploy.md,design.md,develop.md,user-guide.md) to reflect recent framework changes, including the Journal-First philosophy, TCR (Test-Commit-Revert) loop details, and safe update logic.
v0.17.1
Release v0.15.0
Consolidated lifecycle hooks, improved robustness, and enhanced makefile integration.
v0.14.0
[0.14.0] - 2026-03-18
Added
- Notification Hook: Implemented a new
AfterAgenthook (notify.py) that sends a desktop notification usingnotify-sendand plays a system "ping" sound (viapaplay) when the agent completes its turn. - Hook Robustness: Improved project name detection in notifications by correctly identifying the git repository root.
Changed
- Hook Sequencing: Refined the
AfterAgenthook block to be sequential, ensuring notifications only trigger if preceding validation hooks (likemakeorjournal) return anallowdecision.
v0.13.0
[0.13.0] - 2026-03-18
Added
- Hypothesis-Driven Debugging: Refactored the
/debugcommand anddebuggeragent to implement a scientific, hypothesis-driven workflow. Includes status analysis, hypothesis formulation with user approval, and isolated testing in temporarydebug/hyp-*branches. - Content: Added a new blog post draft "Revenge of the Test-Driven Nerds" to the
drafts/directory.
Changed
- Command Rename: Renamed the
.gemini/commands/docs.tomlto.gemini/commands/document.tomlfor better semantic clarity. - Settings: Removed the experimental
plansetting from.gemini/settings.json. - Documentation: Enhanced installation and setup documentation with detailed project scaffolding and integration steps.
v0.12.0
Added
- Implemented a strict TCR (Test-Commit-Revert) protocol for the
/task workcommand, enforcing Red-Green-Refactor development cycles on dedicated feature branches.
Changed
- Refined
install.shto protect project scaffolding files (e.g.,README.md,TASKS.md) from being overwritten while ensuring core framework files are always updated. - Updated documentation and roadmap to reflect the new workflow improvements.