WideCare AI assistant¶
The WideCare AI assistant answers “how do I do this?” from inside WideCare. It knows one thing only: the published WideCare documentation. Every answer shows the pages it comes from.
Ask a question¶
Click the assistant’s icon (a sparkle) in the top bar, available to every user.
Type your question in plain words — the name of the screen or of the step works best.
Press Enter or click the search button.
The window then shows:
the answer — the best passage, with its heading, its page and a link to open the full page; or, when a language model is connected, a written answer;
the other passages — further leads, or the sources of the written answer, each with its link.
When nothing matches, the assistant says so: try other words.
Browse the documentation index¶
WideCare AI → Documentation lists the sections the assistant searches, by language, page and heading, with the link to each page. Search it by word, and group by language or page.
Configure the assistant¶
System administrators open WideCare AI → Configuration (or the Settings app, section WideCare AI):
Answer engine — documentation search only (the default: no key, no network), or a language model, hosted or on a local server, which writes an answer from the same passages and cites them.
Key — for a hosted model, paste the provider’s key and save. It is never shown again; leave the field empty to keep it, or delete it to go back to search only. Save, then use the test button to check it.
Model — leave empty to use the usual model of the chosen engine.
Written answers per day — the limit for the whole database; beyond it, the assistant falls back to the documentation search.
Documentation index — shows the documentation loaded; reload it after an update of the application.
Privacy
The assistant reads only the published documentation. The question asked and the passages found are the only things sent to a language model provider — never a resident’s data.