Search Just Got Smarter — v2.5 Update
v2.5 is a search quality release. The previous versions made LocalSynapse fast and offline. v2.5 makes it smart — it learns from how you actually use it, ranks results more precisely, and handles formats Spotlight has always struggled with.
Phrase search — quote what you mean
Wrap any part of your query in double quotes to require an exact phrase. Search for "quarterly revenue" and you'll only get files where those two words appear together, in that order — no more drowning in files that mention "quarterly" and "revenue" in different paragraphs.
This is the single biggest win for power users who already know what they're looking for. It's a small syntactic addition with a huge precision payoff.
Word boundary boost — cleaner matches first
When you search for report, results where "report" appears as a complete word now rank higher than results where it's just part of another word like "reporter" or "reporting". Subtle, but the top of your results list will feel noticeably more relevant.
Click learning — search adapts to you
v2.4 quietly introduced click tracking. v2.5 turns it into a real ranking signal. When you open a file from search results, LocalSynapse remembers — not just that you opened it, but where it was in the list (position weighting) and whether you came back quickly to the search (bounce detection).
The next time you search for something similar, files you've actually opened in the past rank higher. Files you clicked but bounced off get a smaller boost. The result: your top hits get sharper the more you use the app.
Nothing leaves your machine. The click data stays in the same local SQLite database as the search index.
RTF and ODT now searchable
Previous versions treated .rtf and .odt files as plain text — which mostly worked, but missed structured content and sometimes pulled in formatting markup as if it were prose. v2.5 ships dedicated parsers for both:
- RtfParser — properly extracts text from rich text format, ignoring control codes and formatting noise.
- OdtParser — reads OpenDocument Text files (LibreOffice, OpenOffice) by parsing the underlying XML stream.
If you've got a folder of legacy WordPad .rtf notes or LibreOffice .odt drafts, they're now first-class citizens in search.
Keyboard shortcuts
- Ctrl+K (Cmd+K on Mac) — focus the search box from anywhere in the app.
- Escape — clear the current search query.
Small additions that disappear into your muscle memory after a day.
Empty state, error messages, and download progress
The search page now shows example queries when it's empty, so first-time users have something to try immediately. Errors are written in plain English instead of internal codes. The first-run AI model download (~580MB) shows a progress bar instead of looking frozen. None of these are headline features — they're the kind of polish that makes a tool feel finished.
Detail panel: Size row + no-results tips
Selected files now show file size in the side panel. When a search returns nothing, you get suggestions instead of just a blank screen ("Try fewer words", "Check spelling", "Use semantic search").
Get v2.5
Download from localsynapse.com. If you already have LocalSynapse installed, the search engine improvements apply automatically — no re-indexing needed. The click learning starts from your next search.
Free forever. Apache 2.0. Available for Windows and macOS.