DocuBrowse Files

Deep Links (in-document passage search)

• Fixed semantic Deep Links timing out on large documents (long books/PDFs). Embedding is now bounded, so it returns in seconds instead of failing.
• Relevance floor — semantic Deep Links now returns no passages when a document has nothing genuinely relevant, instead of forcing unrelated "top matches."
• No more junk passages — comment markers (*, //), bare numbers, and other contentless lines are filtered out.
• Cleaner excerpt view — dropped the redundant "return to the main page" note now that an Open full document button exists.

Search

• Semantic search ignores articles and conjunctions (a/an/the, and/or/but/nor/for/so/yet) so short queries match on the words that carry meaning. Keyword search and quoted phrases are unaffected.
• Search runs on Enter, not on every keystroke — no more mid-typing lag or accidental partial-word searches. Clearing the box restores the full list.
• New "Search Tips and Tricks" section in the README: the three modes, "quotes" for exact phrases, and what different word pairings do.

Indexing & housekeeping

• Dotfiles skipped at scan time — hidden files and hidden-directory contents (.env, .git/, .venv/, …) are no longer indexed.
• New purge_dotfiles.py migration tool removes dotfiles indexed by older versions (dry-run by default; ships in all packages).