SD
Public Chat Assistant
Local AI (WebGPU) | optional server proxy
Chat
Local AI (no key)
Server API
Runs fully in your browser using WebGPU.
Load All Models
Only Recommended
Llama-3.2-1B-Instruct (q4) | ~0.6 GB
Qwen2.5-1.5B-Instruct (q4) | ~0.9 GB
Phi-3-mini-Instruct (q4) | ~0.6 GB
Initialize Model
Idle
Auto-speak replies
Test Voice
Stop
Send
Clear
Share
Save
Load
Sizes are estimates from parameters & quantization (q4 ~ 0.5 byte/param) and may vary by packaging.