Curlime is a modern, AI text editor with a beautiful, responsive Electron UI. It leverages LLMs for code generation and provides a safe, interactive playground for running and testing code transforms on your own text.

• AI Code Generation: Generate JavaScript or Python code transforms from natural language prompts using LLMs.
  • Cloud: Anthropic Claude
  • Local (no cloud): Ollama (default)
• Safe Code Execution: Run generated code in a secure Node.js sandbox (vm2) directly from the app.
• Modern UI: Clean, responsive, and beautiful interface with dark code panels, clear labels, and intuitive layout.
• Copy-to-Clipboard: Instantly copy input, prompt, result, or generated code with a single click and visual feedback.
• Preserves Formatting: Output panels maintain original text formatting and support scrolling for large content.
• Full-Screen Experience: The app stretches to fill your entire screen for maximum productivity.
• Electron-based: Cross-platform desktop app with native performance.
• Custom API Key UI: Set and manage your Claude API key from the app.
• Provider Selection: Switch between Claude and local Ollama.
• Health Check & Models: Built-in Ollama health check and available model discovery.

• Save Transformation: Save generated transformations to reuse next time.
• History & Undo: View and restore previous prompts, inputs, and results.
• Export/Import: Save and load sessions or code snippets.
• Drag & Drop: Support for dropping files as input.
• In-app Updates: Notify users of new versions and features.

1. Install dependencies:

   npm install

2. (Optional) Use Claude (cloud):
   • Click on the Settings button in the app UI and save your Anthropic API key
3. (Recommended) Use Ollama locally (no cloud):
   • Install Ollama: see https://ollama.com
   • Start the Ollama server:

     ollama serve

   • Pull a model (default expected model):

     ollama pull deepseek-coder:6.7b

   • You can change the model or server via environment variables (see Configuration below).
4. Run the app:

   npm start

If Ollama is not running, you'll see a warning in the console with quick steps to start it.

• OLLAMA_URL: URL of the local Ollama server
  • Default: http://localhost:11434
• OLLAMA_MODEL: Model name to use with Ollama
  • Default: deepseek-coder:6.7b

Example:

export OLLAMA_URL=http://localhost:11434
export OLLAMA_MODEL=codellama:7b
npm start

• Electron
• Anthropic Claude API
• Ollama (local LLM runtime)
• Node.js (with vm2 sandbox)
• Modern HTML/CSS/JS

Pull requests and suggestions are welcome! See the roadmap above for ideas, or open an issue to discuss new features.

MIT