RunLLM Plugin for PowerToys Run

RunLLM integrates Large Language Models (LLMs) into PowerToys Run, enabling direct AI interaction from the search bar. It supports any service with an OpenAI API-style endpoint (e.g., Ollama, LMStudio).

✨ Features

• 💬 Chat with LLMs: Send prompts and receive streaming responses.
• 🔄 Switch Models: List and switch between available models from the API.
• 🧠 Thinking Mode: Toggle reasoning mode with /think or /no_think.
• 📝 Custom System Prompt: Define custom system instructions.
• 📋 Quick Copy: Copy full or partial responses.
• ⚙️ Configurable: Set API URL, default model, and system prompt.

📦 Requirements

• Any OpenAI-style service (e.g., Ollama, LMStudio).

🔧 Installation

1. Download: Get the release files, including Community.PowerToys.Run.Plugin.RunLLM.dll.

2. Decompress and Copy: Place the files in:
   %LocalAppData%\Microsoft\PowerToys\PowerToys Run\Plugins

   Example structure:

   RunLLM/
   ├── Community.PowerToys.Run.Plugin.RunLLM.dll
   └── Images/
       ├── model.png
       ├── run.png
       ├── change.png
       ├── brain.png
       ├── timer.png
       ├── transfer.png
       └── access.png
   

3. Restart PowerToys: Ensure PowerToys is restarted to load the plugin.

⚙️ Configuration

In PowerToys Settings → PowerToys Run → RunLLM:

• 🌐 LLM URL: Set the API endpoint (e.g., http://localhost:11434 for Ollama).
• 🏷️ Default Model: Specify the default model (e.g., qwen/qwen3-4b, llama3.1:70b, gpt-oss:latest).

🚀 Usage

1. Open PowerToys Run: Alt + Space.
2. Type your query:

   runllm What is the capital of France?
   

📝 Notes

• ✅ If Include in global result is enabled, you can chat without typing runllm.
• 🔑 Use the runllm keyword to change models or toggle thought mode, regardless of global result settings.
• ⚠️ Ensure your LLM service supports /v1/models and /v1/chat/completions endpoints.

🤝 Contributing

Contributions are welcome! Submit issues or PRs at the GitHub Repo.

📜 License

MIT License – see LICENSE.