How do I know whether my assistant is well set up (memory health, entry quality)?
The Memory health card forecasts what the bot will be able to do with your entries. The score is not a grade: it is a forecast, and every missing point has an action to fix it.
The levels
Start › Draft › Solid › Complete › Exemplary. The card shows “N more points to reach …” or “Nothing to fix: the memory is ready”. It also counts the entries and the dictionary terms, and rates the Entry quality and the Server coverage.
The suggested fixes
| Message | What to do |
|---|---|
| N entry/entries the bot mixes up | It will ask which one instead of answering. Make them more distinct (keywords, rewordings). |
| N entry/entries without keywords | Add some (three or more): otherwise they can only be found with the exact words of the question. |
| N entry/entries without rewording | Add one per language, otherwise they cannot be reached from another language. |
| N entry/entries whose answer is still to be written | Replace the text marked ⚠️. |
| N entry/entries added but switched off | Turn them on: they answer nobody. |
| N question(s) left unanswered to review | Real questions the bot couldn’t handle: create an entry for each one. |
Test before turning it on
The simulator replays a whole exchange. Try several wordings of the same question, including with typos and in other languages.
A “Measures” tab
It gathers the assistant’s usage figures. Check it after a few days to see which entries answer well or badly.
Still stuck?
Contact support with the title of this article, the step where you are stuck and what you see on screen.
Contact support More articles: ToolsLast checked: 13 November 2026.