What it does
Live searches combine Jev relevance scores and yes-or-no matches. If no provider initializes, setup selects a keyword-based mock, not Jev inference. Later request failures produce errors, not a switch to mock mode.
Agent tooling
sgrep
Search chunks of code by meaning, then follow a local map of which pieces call other pieces.
Bookmark: sgrep
Keep this in your collection.
Leave a noteWhat would you try with this?
: sgrep
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Start with the README's authentication question over its sample code. Explicit mock mode matches keywords without Jev; live mode sends selected code and the question to the chosen provider. Check which mode produced a result. Exploring related calls requires an existing code map, and local filtering or shortened excerpts can hide relevant evidence.