🎭 omp-model-profiles

Dynamic model profile switcher for Oh My Pi (OMP).

Switch your entire suite of model roles (default, task, slow, plan, vision, advisor, designer, tiny) and thinking levels on the fly without killing, restarting, or leaving your active OMP session.

---

⚡ Features

• 🚀 Live in-session switching: Instantly switch the active model and thinking effort in your current chat.
• 🔄 Full fleet reconfiguration: Applies all model roles (modelRoles) in-memory via OMP's runtime settings override layer so subagents (task), planners (plan), image inspectors (vision), and advisors (advisor) immediately use the new profile.
• 🛡️ Multi-instance isolation: Profile switches only affect the current running terminal session. Other concurrent OMP instances remain completely untouched.
• 🎨 Interactive TUI selector: Type /profile to pop up an interactive dialog with fuzzy search and model previews.
• ⌨️ Direct command shortcut: Fast-switch directly from chat: /profile anthropic, /profile chinese, /profile antigravity, /profile openai.
• 💾 Profile saving: Save your current runtime modelRoles to a reusable profile with /profile save <name>.
• 📦 Zero runtime dependencies: Lightweight, self-contained, and compatible with Bun and Node runtimes.

---

📥 Installation

Single Command (Recommended)

Install directly using OMP's built-in plugin manager:

omp plugin install github:rezhajulio/omp-model-profiles

Or from inside an active OMP chat session:

/plugin install github:rezhajulio/omp-model-profiles

---

🎮 Usage

1. Interactive Picker

Type in your OMP chat:

/profile

(or alias /model-profile)

Opens an interactive dialog showing available profiles and their models:

┌─ Select Model Profile ────────────────────────────────────────────────────────┐
│ > anthropic    default: anthropic/claude-sonnet-4-6:high | slow: ...opus-4-6  │
│   chinese      default: 9router/cmc/deepseek/deepseek-v4.1-flash:high | slow... │
│   antigravity  default: google-antigravity/gemini-3.8-flash:high | slow...   │
│   openai       default: openai/gpt-5.3-codex:auto | slow: openai/gpt-5.4-pro │
└───────────────────────────────────────────────────────────────────────────────┘

2. Switch by Name

/profile anthropic
/profile chinese
/profile antigravity
/profile openai

3. List Profiles

/profile list

4. Inspect a Profile

/profile show chinese

5. Save Current Roles as a Profile

/profile save my-custom-profile

Creates ~/.omp/model-profiles/my-custom-profile.yml.

---

📁 Profile Format

Profiles are stored as standard YAML files inside ~/.omp/model-profiles/<name>.yml.

Example: ~/.omp/model-profiles/antigravity.yml

# Google Antigravity Profile (Gemini 3.8 Flash, Gemini 3 Pro)
modelRoles:
  default: google-antigravity/gemini-3.8-flash:high
  smol: google-antigravity/gemini-3.8-flash:low
  slow: google-antigravity/gemini-3-pro:high
  vision: google-antigravity/gemini-3.8-flash:low
  plan: google-antigravity/gemini-3-pro:max
  task: google-antigravity/gemini-3.8-flash:low
  advisor: google-antigravity/gemini-3-pro:high
  commit: google-antigravity/gemini-3.8-flash:low
  designer: google-antigravity/gemini-3-pro:high
  tiny: google-antigravity/gemini-3.8-flash:low

Thinking Level Suffixes

You can append :thinkingLevel to any model selector:

• :none — disable reasoning/thinking tokens
• :low — minimal thinking budget
• :medium — balanced reasoning
• :high — deep reasoning
• :max — maximum reasoning effort
• :auto — provider default

---

📚 Example Profiles Included

Check the examples/ directory for ready-to-use profiles:

Profile	Primary (default)	Reasoning (slow / plan)	Subagents (task)	Best For
anthropic.yml	Claude Sonnet 4.6	Claude Opus 4.6 / Opus 5	Claude Haiku 4.5	Deep architecture, refactoring & code review
chinese.yml	CMC DeepSeek V4.1 Flash	CMC DeepSeek V4 Pro	CMC DeepSeek V4.1 Flash	High speed, 1M context & low token cost via 9router
deepseek.yml	DeepSeek V4.1 Flash	DeepSeek V4 Pro	DeepSeek V4.1 Flash	Direct native DeepSeek provider integration
antigravity.yml	Gemini 3.8 Flash	Gemini 3 Pro	Gemini 3.8 Flash	Fast loops, native vision & 1M context
google.yml	Gemini 3.8 Flash	Gemini 3.1 Pro Preview	Gemini 3.8 Flash	Direct Google AI provider with large context
openai.yml	GPT-5.3-Codex	GPT-5.4-Pro	GPT-5.4-Mini	Advanced reasoning & frontier coding
xai.yml	Grok 4 Fast	Grok 4.20 Reasoning	Grok 4 Fast	High-speed xAI coding with deep reasoning
local-ollama.yml	Qwen2.5-Coder 32B	DeepSeek-R1 32B	Qwen2.5-Coder 7B	100% offline & local execution

To install all example profiles:

mkdir -p ~/.omp/model-profiles
cp examples/*.yml ~/.omp/model-profiles/

---

🛠️ Development & Testing

# Run test suite (Bun test runner)
bun test

# Run tests with coverage
bun test --coverage

# Bundle single-file distributable
bun run build

---

👤 Author

Rezha Julio

• Email: contact@rezhajulio.id
• GitHub: @rezhajulio

---

📄 License

MIT License © 2026 Rezha Julio