Search by meaning
- Spotlight
- Limited
- Cloud AI
- Yes
- Aether
- Yes (on-device)
What is Aether
A local-first AI file search app that finds documents by meaning.
Aether is local semantic file search for macOS.
You point it at folders on your Mac; it builds an on-device index with vector embeddings and keyword signals.
Then you search like you think: “that PDF about the board deck,” not only board-deck-final-v3.pdf.
Unlike cloud AI document search, Aether never uploads your library. Unlike Spotlight alone, it is tuned for idea- and phrase-based retrieval across PDFs, Markdown, Office files, email, images (OCR), and source code.
The problem
You remember the idea. Spotlight remembers the name.
macOS Spotlight is excellent when you know the filename. It falls apart when you remember a phrase, a topic, or “the notes from that client call.” And most cloud AI search tools require uploading private files to someone else’s servers.
Aether fills that gap: private document search with hybrid semantic + keyword ranking, a Spotlight-style launcher, and a full workspace, built for researchers, developers, writers, and anyone drowning in local PDFs and project folders.
How it works
Add folders. Index once. Search anytime on your Mac.
Aether is a native SwiftUI macOS app paired with a local Python search engine (LanceDB + sentence-transformers). You choose what to index. Embeddings and ranking stay on-device.
-
01
Add folders to index
Pick project directories, Documents, Downloads: whatever you actually search. Drag and drop works. Files stay in place; Aether only reads them to build a local search index.
-
02
Index with on-device embeddings
Content is chunked and embedded with a local model (all-MiniLM-L6-v2). Progress shows in the app; cancel anytime. The vector index lives in ~/.aether on your disk.
-
03
Search by meaning or keywords
Press ⌘⇧Space for the launcher, or open the workspace. Type an idea, phrase, or filename, then open, preview, or Quick Look the right file.
Compare
Spotlight vs cloud AI search vs Aether
People searching for a “Spotlight alternative,” “private AI file search,” or “offline Mac document search” usually need one of three things. Here’s how Aether fits.
Files stay on your Mac
- Spotlight
- Yes
- Cloud AI
- Usually no
- Aether
- Yes
Works offline after setup
- Spotlight
- Yes
- Cloud AI
- Rarely
- Aether
- Yes
PDF / OCR / code content
- Spotlight
- Partial
- Cloud AI
- Varies
- Aether
- Built for it
Account & telemetry
- Spotlight
- System
- Cloud AI
- Typically required
- Aether
- None
| Capability | Spotlight | Cloud AI search | Aether |
|---|---|---|---|
| Search by meaning / semantics | Limited | Yes | Yes (on-device) |
| Files stay on your Mac | Yes | Usually no | Yes |
| Works offline after setup | Yes | Rarely | Yes |
| PDF / OCR / code content | Partial | Varies | Built for it |
| Account & telemetry | System | Typically required | None |
Features
Everything you need for private Mac file search without the cloud.
Aether is not a thin chat wrapper. It is a full local search product: launcher, workspace, indexing controls, OCR, related files, and duplicate detection designed for daily use.
Hybrid semantic + keyword search
Queries are embedded with a local model, then ranked with filename matches, path signals, content keywords, and light synonym expansion, so “invoice” can still surface a receipt, while exact names stay on top when they should.
Scope search to one folder or your whole library. Filter by documents, code, data, or images. Sort by relevance, name, or date.
Spotlight-style launcher & workspace
The launcher is a compact overlay for fast open-and-go search: keyboard navigation, recent queries, saved searches, and ⌘↩ to jump into the workspace with preview.
The workspace gives you three panes (folders, results, and preview) for deeper browsing. Select a hit to read a snippet with query highlighting, jump to related files, or press Space for native Quick Look.
Broad file & OCR support
Index PDFs, Word, Pages, text, Markdown, spreadsheets, slides, email (.eml), and common source files: Swift, Python, TypeScript, HTML, Rust, Go, and more.
Images (PNG/JPEG) go through on-device OCR. Scanned PDFs fall back to OCR when there is no extractable text layer, so screenshots and scans can still be found.
Indexing you control
Index one folder or everything. Re-index when you need a clean pass. Background indexing keeps search available while work runs. Cancel anytime. Status shows progress and how many files were skipped when extraction fails.
Optional live sync watches folders for changes. Settings show index health per folder: file counts, chunk counts, and last indexed time, plus exclude rules for paths you never want scanned.
Related files & duplicates
Open a result and Aether suggests related files from the same semantic neighborhood, useful when the right note is near the one you found, not identical to your query.
Duplicate detection groups exact copies and near-duplicates from the index, so you can clean up sprawling Downloads folders without uploading anything.
Privacy by architecture
No account. No telemetry. No cloud sync of your corpus. The embedding model runs on your Mac. Index data stays under ~/.aether.
A menu bar item keeps Aether one click away. Onboarding and Settings stay honest about what is local, including the one-time model download on first launch.
Supported types
Built for the files already on your Mac.
Documents
PDF, DOCX, Pages, TXT, Markdown, EML
Office & data
XLSX, PPTX, CSV, JSON, LOG
Code
Swift, Python, JS/TS, HTML/CSS, Rust, Go, and more
Images
PNG, JPG, JPEG with on-device OCR
FAQ
Questions people ask about local semantic search
Straight answers for humans, search engines, and AI assistants citing what Aether is and is not.
What is Aether?
Aether is an open-source, local-first semantic file search application for macOS. It indexes folders on your computer and lets you find documents by meaning, keywords, or filename, without uploading files to the cloud.
How is Aether different from macOS Spotlight?
Spotlight is excellent when you know a filename or metadata. Aether is built for when you remember an idea, phrase, or topic. It combines on-device semantic embeddings with keyword and filename ranking so you can search the content of PDFs, notes, and code by meaning.
Does Aether upload my files to the cloud?
No. Indexing, embeddings, and search run on your Mac. There is no account, no telemetry SDK, and no cloud sync of your document corpus. Index data lives in ~/.aether on disk. The only network step is the optional first-time download of the public embedding model weights.
What file types can Aether search?
PDFs, Word (DOCX), Pages, text, Markdown, spreadsheets (XLSX), slides (PPTX), email (.eml), common source code formats, and images (PNG/JPEG) via on-device OCR. Scanned PDFs can fall back to OCR when no text layer exists.
Is Aether free and open source?
Yes. Aether is MIT-licensed and unfinished on purpose: clone it, improve it for your workflow, or open a pull request. Run ./Scripts/setup-backend.sh, then build with Xcode.
Does semantic search work offline?
After the first-time download of the local embedding model (all-MiniLM-L6-v2), search and indexing run offline using the cached model and your local LanceDB index.
Why I build this
Private search should not require a cloud account.
I am building Aether because remembering a filename should not be the price of finding your own work. Spotlight is great for names. Cloud AI search is great for meaning, until it asks you to upload private PDFs, notes, and code. I want the middle path: semantic search that stays on your Mac.
This is also a bet on local-first software. Your documents are already on disk. The index should live next to them, not in someone else’s product account, with a subscription and a privacy policy you have to trust.
Honest take: Aether is not perfect.
Ranking will miss. Some file types are rough. First setup asks you to run a backend script. Packaging a one-click DMG with a bundled Python runtime is still future work. That is fine for an open-source project: ship something useful, keep it local, and get better in public.
If Aether helps you, fork it and shape it for your folders. If you fix a bug, improve retrieval, or clarify docs, contribute back. Issues and pull requests are welcome. MIT license means you can also take it and build your own direction.
Get started
Keep your files. Find them by meaning anyway.
Clone the repo, run the backend setup once, build from Xcode. First launch caches the search model; later launches stay on your machine. Then search, break things, improve it, and share what you learn.