CFR Find

Fast, worker-powered full-text search for Obsidian.

Credit where it's due: CFR Find is heavily inspired by Omnisearch by Simon Cambier, licensed under GPL-3. Omnisearch pioneered this style of ranked, typo-tolerant vault search in Obsidian, and CFR Find borrows many of its ideas (and its Text Extractor integration). If you need an HTTP API, a public plugin API, embed results, or Chinese word segmentation, use Omnisearch — it's excellent. CFR Find trades those features for raw speed. Like Omnisearch, CFR Find is licensed under GPL-3.

Why another search plugin?

Omnisearch's slowness on large vaults is architectural: all indexing, tokenization, and cache serialization run on Obsidian's UI thread, every document is tokenized several times over, and the whole index is serialized as one blob on the main thread. CFR Find keeps the same core idea (BM25 ranked search via MiniSearch) but restructures everything around performance:

	CFR Find	Omnisearch
Indexing & search thread	Web Worker (UI never blocks)	Main/UI thread
Tokenization	Single pass, deduplicated	3–4 passes, duplicates kept
Diacritics normalization	Once per distinct token (memoized)	Per document + per term + per query
Note content in memory	Not retained	2 full copies of every note
Cache serialization	In the worker, one atomic transaction	On the UI thread
Excerpt highlighting	Visible results only (IntersectionObserver), offset-based <mark>	All 50 results, regex replace
UI	Plain DOM, no framework	Svelte 5
Bundle	~75 KB	~1 MB+

Features

• Vault-wide search (CFR Find: Search the vault): ranked, typo-tolerant, prefix-matching search across markdown, canvas, and any plaintext extensions you configure.
• In-file search (CFR Find: Search in the current note): jump to any match in the active note. Tab switches between the two modes, carrying your query.
• Query syntax: "exact phrases", -excluded words, -"excluded phrases", ext:md / .md extension filters, path:Daily / -path:Archive filters.
• Smart matching: diacritics-insensitive (cafe finds café), camelCase-aware (user finds getUserName), hyphen-aware.
• Instant restarts: the index is persisted (IndexedDB) and only files whose modification time changed are re-indexed on launch.
• Live index: created/renamed/deleted notes update immediately; edits are re-indexed after a 2-second pause and always flushed before a search opens.
• Configurable fuzzy search: typo tolerance level (off → aggressive), minimum word length for fuzzy matching, and prefix-matching threshold — all applied instantly, no re-index needed.
• PDF / image (OCR) / Office indexing via the Text Extractor community plugin. Disabled by default — extracting text is slow and can slow down Obsidian during indexing; a warning is shown when you enable it.
• Field-weighted ranking: filename and aliases > tags > headings > body.
• Respects Obsidian's Excluded files setting.
• Custom icon and a keyboard-first UI with a fixed-size result panel.

What CFR Find does not do

Being honest about scope — these Omnisearch features are intentionally not in CFR Find (use Omnisearch if you need them):

• No public API for other plugins and no obsidian:// URL scheme.
• No local HTTP server.
• No embed results (images/documents surfaced under the notes that embed them).
• No Chinese word segmentation integration.
• PDF/image/Office indexing requires the separate Text Extractor plugin (desktop only, same as Omnisearch) and is off by default because text extraction is slow.

Opening the search

The command is CFR Find: Search the vault (there is also CFR Find: Search in the current note). Bind a key to it in Settings → Hotkeys, or use the shortcut button in CFR Find's settings tab which takes you straight there.

Install (manual)

1. Run npm install && npm run build.
2. Copy dist/main.js, dist/manifest.json, and dist/styles.css into <your vault>/.obsidian/plugins/cfr-find/.
3. Reload Obsidian and enable CFR Find in Community plugins.

Development

npm install
npm run dev    # watch build
npm test       # unit tests (tokenizer, query parser, engine, worker core)
npm run build  # type-check + production build

Architecture notes: the search engine runs in a Web Worker that is bundled separately and inlined into main.js as a blob URL (Obsidian plugins ship a single file). The main thread only reads files and renders UI; file contents are posted to the worker and never kept resident. See src/shared/protocol.ts for the full message contract.

Privacy and network use

CFR Find makes no network requests. All indexing and searching happens locally on your device; the index is stored locally in IndexedDB. The only external link in the plugin is the optional "Buy me a coffee" button in the settings tab, which opens buymeacoffee.com in your browser only if you click it.

Support

If CFR Find is useful to you, you can buy me a coffee. ☕

Credits

• Omnisearch by Simon Cambier — the inspiration for this plugin and the pioneer of ranked fuzzy search in Obsidian.
• MiniSearch by Luca Ongaro — the BM25 full-text search engine CFR Find runs inside its worker.
• Text Extractor by Simon Cambier — powers the optional PDF/image/Office indexing.

License

GPL-3.0-or-later. Inspired by Omnisearch © Simon Cambier.