Native Search

English | 日本語

Zero-index search for Obsidian, powered by the index your OS already maintains.

Search the full text of PDFs, the OCR'd text inside images (via Apple Live Text), Office documents (Word, PowerPoint, Keynote, Excel, …) and your notes — with zero indexing cost inside Obsidian. There is no plugin-side background index, cache file, index-related startup delay, or resident search index. The plugin-side footprint stays nearly flat even for vaults with 100,000+ files.

Why

An in-app search engine must build, load, update, and retain its own index in Obsidian's renderer process. Those costs are often negligible for an ordinary vault, but they grow with the number and size of indexed files.

macOS Spotlight already maintains a full-text and OCR index of your vault folder. Native Search queries that existing index with mdfind, leaving indexing and extraction outside Obsidian. This architecture is especially useful for very large or document-heavy vaults.

Native Search trades relevance ranking and typo tolerance for a minimal in-app footprint. See the comparison below for the practical differences.

Usage

• Click the ribbon icon, or run the command "Open Native Search" (assign a hotkey for the full Omnisearch-style experience).
• Type any words — multiple words are AND-combined. Matching is case- and diacritic-insensitive, so poincare can find Poincaré. Japanese, English and other languages are supported wherever Spotlight indexes them.
• Use "exact phrase" for a phrase, -word or -"excluded phrase" to exclude content and filename matches, path:Projects/ for a vault-relative path prefix, and ext:pdf for a file extension.
• Results show badges: NOTE / PDF / IMAGE (OCR hit) / OFFICE / FILE, plus NAME for filename matches (shown first).
• Results are grouped NAME → NOTE → PDF → IMAGE → OFFICE → FILE; files within each group are newest first. Configurable excerpts show the first matching context for text-based files.
• Reopening the modal restores the previous query and results within the current Obsidian session. The query is selected so typing replaces it immediately.
• ↵ opens in the current tab, ⌘↵ in a new tab. Files Obsidian can't display (e.g. .docx, .key) open in their default app.
• Opening a note jumps to the first match and highlights every occurrence — the same mechanism Obsidian's core Search uses. Opening a PDF drives the built-in viewer's search bar: all matches are highlighted and the view jumps to the first one. (Images can't be highlighted in place — Spotlight doesn't expose OCR match positions.)

Settings

• Max results / Debounce
• Include notes — turn off to search only non-note files (use alongside your favorite note search)
• Search filenames — filename matches ranked first
• Show excerpts / Excerpt count — show matching context for up to the configured number of top text-file results
• Remember last query — restore the previous query and results when reopening the modal
• Excluded folders — vault-relative prefixes, one per line

Comparison with Omnisearch

Omnisearch is a mature, excellent in-app search engine. It and Native Search optimize for different constraints:

Capability	Native Search	Omnisearch
Relevance ranking (BM25)	No; deterministic match/type groups, then modification time	Yes; weighted BM25 document scoring
Fuzzy matching	No	Yes; typo-resistant search
Excerpts	Yes, for a configurable number of text-file results	Yes
Phrase & operators	Quotes, exclusions, path: prefixes, ext: filters	Quotes, exclusions, extension filters
PDF full-text	Spotlight metadata index	Via the companion Text Extractor plugin
Image OCR	Spotlight / Apple Live Text	Via the companion Text Extractor plugin
Office documents	Spotlight metadata importers	Via the companion Text Extractor plugin
Indexing cost	No index inside Obsidian; the OS maintains it	Builds and updates a MiniSearch index inside Obsidian
Startup impact	No plugin index to load or rebuild	May load or update its cache/index; cost depends on the vault
Memory footprint	No resident plugin search index	Resident index grows with the indexed corpus
Vault-size scaling	Near-flat plugin-side cost; particularly suited to 50k–100k+ files	Indexing, cache, and memory work grow with the corpus
Index freshness	Spotlight updates asynchronously; timing is controlled by macOS	Vault changes can be reflected immediately
Platforms	macOS desktop only	Desktop and mobile (dependency features may vary)

For small to medium vaults, Omnisearch is an excellent choice: its relevance ranking, fuzzy matching, broad platform support, and immediate index updates make it a more capable general-purpose search experience.

Native Search's advantage becomes clearest in huge vaults. In one 100k-file stress test, an in-app indexer added about 9 seconds to startup, used roughly 668 MB of additional heap, and could freeze the renderer while writing its cache. Native Search added no plugin-side index build, cache write, or resident index. Those figures are a workload-specific observation rather than a universal benchmark, but they illustrate the architectural difference.

Requirements & limitations

• macOS only. Spotlight (mdfind) is a macOS facility. (Windows Search / Linux equivalents are conceptually possible — PRs welcome.)
• Your vault must live in a Spotlight-indexed location (most home-folder locations are). Check with:

  mdfind -onlyin /path/to/vault "kMDItemTextContent == '*someword*'cd"
  

  If this returns nothing for a word you know exists, check System Settings → Siri & Spotlight → Spotlight Privacy, and note that some external/network drives are not indexed.
• Spotlight updates are asynchronous. A just-created or just-edited file may take time to appear, and Native Search cannot force or rebuild the OS index.
• Image OCR coverage is whatever Apple Live Text has processed — it covers common image formats and improves over time, but is controlled by the OS, not the plugin. Spotlight reports that an image matched, not where, so the match can't be highlighted inside the image.
• PDF match highlighting drives the built-in viewer's find bar, which is not a public API; if a future Obsidian version changes the viewer internals, the PDF still opens and only the automatic highlight is skipped.
• Excerpts are read only for the configured number of top text-file results. PDFs, images, and binary Office formats do not receive excerpts.
• Spotlight does not expose relevance scores via mdfind, so results are grouped (filename → notes → PDF → images → Office → other files) rather than relevance-ranked.
• Obsidian's "Excluded files" setting is not consulted; use the plugin's own excluded-folders setting.

Privacy

Everything runs locally. The plugin spawns mdfind (a standard macOS command-line tool) against your vault folder only. No network requests, no telemetry, no index files written.

Install

Until the plugin is accepted into the community directory:

1. Download main.js, manifest.json, styles.css from the latest release.
2. Put them in <vault>/.obsidian/plugins/native-search/.
3. Enable Native Search in Settings → Community plugins.

Or use BRAT with gyroid-eth/obsidian-native-search.

License

MIT