Codelves

Three panels

☰ FilesOPFS file system — persists in your browser. Create, open, delete files and folders. EditorWrite Python (or any text). Open files appear as tabs. Unsaved changes shown with ● ✦ LLMAsk your local Ollama or Gemini API about the open file. Toggle with the LLM button.

Keyboard shortcuts

Ctrl+EnterRun current file with Pyodide Ctrl+SSave file to OPFS Ctrl+JToggle LLM chat panel TabInsert 4 spaces

LLM setup

LM StudioLoad any GGUF in LM Studio → start local server → copy model name into ⚙ Settings. Port 1234. OllamaRun ollama serve locally. Default model: qwen2.5-coder:7b. No API key needed. GeminiPaste your API key in ⚙ Settings. Uses gemini-2.0-flash by default. wllamaRecommended. Pick any .gguf — cached in IndexedDB after first load. Real token streaming. Works with LM Studio downloads. transformers.jsDownloads ONNX model from HuggingFace (~300MB). No file needed, but slower and heavier than GGUF. Inject fileCheckbox in LLM panel sends the open file as context with every message.