Why does the local AI need WebGPU?
The Research Assistant's AI runs on your own computer instead of on a remote server. It uses WebGPU, a browser feature that lets web pages use your graphics chip, to run a language model fast enough to be useful.
Why it's built this way
Running the model locally means your prompts and project text are processed on your device and aren't sent to an AI service. The trade-off is that your browser and hardware have to support WebGPU.
Which browsers support it
The app recommends:
- Chrome 113 or later
- Microsoft Edge 113 or later
- Opera 99 or later
It lists Firefox 118 or later and Safari 17 or later as experimental. Even in a supported browser, WebGPU may be unavailable if your graphics hardware or drivers don't support it, or if the graphics chip doesn't have enough memory for the model.
If WebGPU isn't available
The Research Assistant shows a notice and keeps working in a basic mode with rule-based help. Your projects, tools, and exports all still work. Only some AI features are limited or unavailable.
How to turn the AI on
The AI is off until you choose to enable it. Enabling it starts a one-time model download. See Why is the first AI load slow?