Releases: code-yeongyu/oh-my-opencode
v1.1.0
New Features
🔍 look_at Tool & multimodal-looker Agent
- Add
look_attool for analyzing media files (PDFs, images, diagrams) - Add
multimodal-lookeragent using Gemini 2.5 Flash model - Analyze visual content in separate context window, saving main conversation tokens
Inspired by Sourcegraph Ampcode's look_at tool design.
Usage
look_at({
file_path: "/path/to/diagram.png",
goal: "Describe the architecture components"
})Supported File Types
- PDF documents
- Images (PNG, JPG)
- Screenshots
- Architecture diagrams
- Charts/Graphs
v1.0.2
v1.0.1
- 55bc8f0 refactor(ultrawork-mode): use history injection instead of direct message modification
- 0ac4d22 feat(think-mode): inject thinking config with maxTokens for extended thinking
- 19b3690 docs: add Ultrawork Mode hook documentation
- 564c8ae fix: use
lstatSyncinstead ofstatSyncfor symlink detection (#32)
v1.0.0
v1.0.0
The first stable release of oh-my-opencode.
Every feature I envisioned is now implemented. Every integration I dreamed of is now real. After extensive testing and stabilization, I can finally say with confidence: this is production-ready.
This plugin is your secret weapon. While it is open source, those who install it will be one step ahead of the developer sitting next to them. The multi-model orchestration, the LSP tools, the AST-Grep search, the background task system, the automatic context injection - all working together to amplify your capabilities beyond what any single-model agent can achieve.
What's New Since v0.4.3
Features
- Ultrawork Mode Hook: Trigger "ultrawork" or "ulw" to activate maximum agent orchestration. Forces parallel exploration, strict TODO tracking, and full utilization of all specialized agents via background tasks.
- Disabled Hooks Configuration: Selectively disable any built-in hook via
disabled_hooksconfig option. - Background Task Improvements:
- Show original prompt and last message in running task status for better visibility
- Send notifications to parent session instead of main session for correct targeting
- GCP Permission Error Retry: Exponential backoff for Google Cloud permission errors in Antigravity auth - no more manual retries.
Bug Fixes
- Fixed session-notification race conditions and memory issues
- Improved background-task notification targeting
Documentation
- Updated LLM agent guide with Google Auth recommendation
- Added
disabled_hooksconfiguration section
Contributors
Thank you to everyone who contributed to this release:
v0.4.4
Features
- Configurable hooks: Add
disabled_hooksconfig option for selectively disabling hooks (#28) - Background task status: Show original prompt and last message when checking running task status
- GCP retry logic: Add exponential backoff retry for Antigravity GCP permission errors
Fixes
- Session notification: Fix race condition and memory issues in session-notification hook
Documentation
- Update README with Google Auth recommendations and disabled_hooks configuration section
v0.4.3
Release Notes: v0.4.3
Since: v0.4.0 (2025-12-12)
🔧 Improvements
Google Auth Integration (Breaking Change from v0.4.0)
Before (v0.4.0):
{ "plugin": ["oh-my-opencode", "oh-my-opencode/google-auth"] }After (v0.4.2+):
// opencode.json
{ "plugin": ["oh-my-opencode"] }
// oh-my-opencode.json
{ "google_auth": true }The Google Antigravity OAuth plugin is now integrated into the main package. Enable it via google_auth: true in your oh-my-opencode.json configuration.
This change resolves #30 - OpenCode's plugin loader was incorrectly treating oh-my-opencode/google-auth as a separate npm package.
📚 Documentation
- Add Google Auth section to Configuration in README.md / README.ko.md
- Update JSON schema with
google_authandlspoptions for IDE autocomplete
🐛 Bug Fixes
v0.4.2
v0.4.0
Release Notes: v0.4.0
Since: v0.3.4 (2025-12-12)
Commits: 53 | Files: 58 | Changes: +7,314 / -116
✨ New Features
🔐 Antigravity Auth (Google Gemini Support)
Built-in OAuth authentication plugin for Google Gemini models.
{ "plugin": ["oh-my-opencode", "oh-my-opencode/google-auth"] }- OAuth 2.0 with dynamic port allocation
- True SSE Streaming implementation
- Thought Signature - multi-turn conversation tracking
- Request/Response transformation (OpenAI ↔ Gemini)
- Tools + Extended Thinking support
🚀 Background Agent System
Asynchronous task execution system.
background_task/background_output/background_canceltools- Auto-completion notification to main session
- Memory-only persistence
📜 Rules Injector
Support for .claude/rules/ and .cursor/rules/ directories.
- Frontmatter
globspattern matching alwaysApply: trueoption.mdand.mdcfile support
🔄 Auto Update Checker
Automatic plugin version checking with update notifications.
🤖 call_omo_agent
Subagent orchestration tool for spawning explore and librarian agents.
🔧 Improvements
- anthropic-auto-compact: Exponential backoff retry mechanism
- context-window-monitor: Fixed token calculation method
- session-recovery: Handle empty messages + API/Storage index mismatch recovery
🐛 Bug Fixes
antigravity: Use loadCodeAssist project IDantigravity: Improve streaming retry logicbackground-agent: Multiple notification bug fixes (17 commits)
♻️ Refactoring
- Remove
opencode-openai-codex-authdependency - Rename
omo_task→call_omo_agent - Unify
background_result/status→background_output
📚 Documentation
- Add Background Task section
- Improve authentication setup guide with detailed provider instructions