Hey HN, I built Rememex a semantic search layer for your local files. The problem: I kept losing files. Not because they were deleted, but because I couldn't remember the exact filename or keyword. grep needs the exact word. Everything only searches filenames. I wanted to type what I meant and find what I needed. How it works: - Indexes 120+ file types (code, docs, images, configs) - Hybrid search: vector embeddings + full-text + JINA cross-encoder reranking - OCR on images via Windows UWP engine - Reads EXIF GPS → reverse geocodes to city names ("photos from istanbul" works) - EXIF dates → human language ("summer morning" finds a July 8am photo) - Smart chunking per language (Rust at fn/struct, Python at def/class) - Built-in MCP server so AI agents can use it as a tool Everything runs locally. Embeddings use a local ONNX model (Multilingual-E5-Base) by default, though you can optionally plug in OpenAI/Gemini/Cohere. Named after Vannevar Bush's Memex (1945) his vision of a device that stores and retrieves all human knowledge. Stack: Rust (Tauri 2), React/TypeScript, LanceDB, rayon I benchmarked it against grep for agentic tasks rememex consistently finds things in 1 step where grep takes 3-5 or fails entirely. The key difference: grep needs the exact keyword, rememex needs the idea. Windows-only for now (UWP OCR dependency), but the core engine is portable. Would love feedback on the search quality and architecture. MIT licensed, free forever.