Local Intelligence
Fast responder active (default)
Network on. A download starts only after you confirm it.
This switch applies to Lattice Chat models only.
Third-party models under their own licenses. Not made by Ra-Thor. Not reviewed or endorsed by their authors.
The first download of each model comes from Hugging Face and GitHub (raw.githubusercontent.com, which serves the model's code file) and needs the network. After that it runs in this browser.
A small size file is fetched to show the download size.
Any other model: Local Server (Ollama).
Pick a preset, then Connect. You can edit both fields.
WebLLM = pure browser. Local Server = Ollama / LM Studio / any OpenAI-compatible endpoint on your machine.
Local session store • Multi-session • Optional passphrase (PBKDF2 + AES-GCM) • No backend we control
Bridge to any cloud LLM
Recommended on phones: copy a clean system prompt + history, then paste into Grok, Claude, Gemini, ChatGPT, or any model.