There is a globe next to the input box in the assistant.
- It switches on for one question. Press the globe and that question goes out with a web search behind it. The next one answers normally unless you press it again. The mode does not stick, so you cannot leave it on for a whole conversation by accident.
- The answer comes with its sources. Underneath it, a collapsed list of the links it was built from. One click opens it, and it takes no room in the thread meanwhile.
- Use it for what the model cannot know. A rate, a price, what shipped this week, whether that offer still stands. An ordinary answer is limited to what the model learnt during training; search lifts that limit exactly when you need it lifted.
What leaves your machine. The text of your question goes not only to the AI provider but to a search service as well — there is nothing to search otherwise. The principle is the one you already know: cloud features are switched on one at a time and only by you, and your ordinary typing is never sent anywhere at all. It is just that here there are two recipients, and you should know that before you press the globe rather than after.
It works with NeuroSwitcher Cloud only. The search lives on our proxy, which decides how to search and what it costs. If your assistant is set to your own OpenAI key, a custom endpoint or a local model, there simply is no globe next to the input box — that is a capability those backends do not have, not a fault. Switch the assistant to NeuroSwitcher Cloud in the AI assistant tab and the button appears.
About the counter. Searches are counted separately from ordinary questions, and there are noticeably fewer of them. That is arithmetic, not meanness: one search costs about what a thousand plain questions to the model cost. How many are left today is written under the input box, next to the other allowances.
To update: the console will offer the update itself, or download the installer from the site. It installs over your current copy — settings, dictionaries and licence are kept.