Sable Concierge
Before we begin

This concierge runs a real neural network in your browser.

No servers, no transcripts kept — your questions never leave this device. The model is fetched once, then the desk answers from memory, even with the connection gone.

Reaching for the model…0%

all-MiniLM-L6-v2, quantized · fetched from the Hugging Face CDN

Semantic retrieval over 28 notes · all-MiniLM-L6-v2 · runs on WebGPU/WASM in your device desk closed

Want something like this for your brand?

Start a project → See more work