Why a note you know exists does not turn up in search

There are three separate searches in NeuralVault, not one, and only the one the assistant runs on its own behalf is genuinely hybrid. The other two are plain substring matches.

Last checked against the app on · View as Markdown

Steps

  1. The Library page itself has no search box. It is folder browsing: open a folder, open a note.
  2. The @ picker in the chat composer matches against note TITLES only, as a case-insensitive substring, and needs at least two characters before it searches at all. A phrase inside a note but not in its title will not surface it.
  3. Searching past conversations, from the chat history list, matches both the title and every message’s content, which is a different search again from the note picker above.
  4. True hybrid search, lexical and semantic together, only runs when the assistant itself searches the vault as part of answering you. There is no button that runs it directly.

How to tell it worked

A note you can find by opening its folder by hand is the note you were looking for. That confirms nothing was lost; it tells you which of the three searches was the wrong tool for the job.

If it does not work

You typed a phrase you know is inside a note, using @ in chat, and got nothing.
The @ picker only reads titles. Rename the note to include the phrase, or open it directly from the Library instead.
The assistant says it searched your vault semantically, but answers as if it only matched keywords.
Semantic search needs a reachable embedding model. If Ollama is offline, times out, or the embedding call fails for any reason, the search silently falls back to keyword-only, with no error shown to you.
You want one search box that reaches everything.
There is not one in this build. Which of the three you want depends on whether you are looking for a note, a past conversation, or asking the assistant to find something for you.