Find any file by what it means
Filesy is a private, on-device semantic search for macOS. Index your local directories once, then search in Russian, German, or English. Everything runs entirely on your machine using Qdrant vector database, Ollama with BGE-M3 multilingual embeddings and Qwen 2.5 for query cleanup, and a local fastembed Cross-Encoder reranker. Fully private, zero cloud, zero telemetry. Built-in Swift Apple Vision OCR helper for scanned documents.
Multilingual Semantic Search
Find PDFs, documents, notes, and code by intent - not just exact filenames. Powered by Ollama BGE-M3 1024-d embeddings and local Qwen 2.5 query cleanup, supporting Russian, German, and English queries.
Fully Local & Air-gapped
Runs locally on your Mac using Qdrant (Docker) and Ollama. Your files, vectors, query logs, and index inventories stay strictly offline in your device environment. Zero external telemetry.
Hybrid Retrieval & Apple OCR
Combines raw and cleaned query embeddings so key terms are never missed. Features local Cross-Encoder reranking via Jina-v2 fastembed ONNX (CPU) and a lightning-fast native Apple Vision OCR swift-binary for scanned PDFs.
How it works
Index directories & automate sweeps
Point Filesy at folders in the desktop Cockpit. It walks directories, filters whitelists, computes sha256 hashes to skip unchanged files, and schedules sweeps via APScheduler.
Type queries in plain language
Enter questions or describe what you remember. The background FastAPI daemon cleans syntax, embeds the hybrid queries, and retrieves highly relevant chunks.
Review in five-tab Tauri app
A local Tauri 2 + Svelte 5 desktop app offers Search, Folders, History, Cockpit, and Manual tabs. View exact matching passages and open files instantly without overhead.