Real model inference
Responses come from an open-weight model running in the browser, not from a library of scripted replies.
Meet DECAN AI — an experimental assistant exploring open models, useful memory, and knowledge gathered from permitted public sources.
DECAN AI is a live experiment. It starts small, measures what works, and makes room for stronger infrastructure as the project grows.
Responses come from an open-weight model running in the browser, not from a library of scripted replies.
A scheduled collector fetches a small number of permitted public feeds and records their source links.
New knowledge can be indexed without changing model weights. Model training is a separate future research milestone.
DECAN AI currently runs its model in the visitor's browser. Initial model downloads may be large, WebGPU support varies by device, and free hosting/storage limits apply. Online collection does not mean the model retrains itself every second.