Recall: Open-Source Local Semantic File Search Hits 95% Accuracy in 0.17s on an i5 Laptop
somthing_tn · reddit · 2026-10-11
A developer released Recall (MIT license), a Rust-based local semantic file search tool for Linux: it indexes PDFs, Office files, scans, screenshots, audio and video, and retrieves by content — a query like "when did they decide to freeze hiring" opens the recording at that second.
Technical details
- EmbeddingGemma 2 embeds text, images, video frames and 30s audio windows into one vector space, run in-process via llama.cpp (CUDA, Vulkan or CPU).
- Hybrid ranking: vector similarity plus SQLite FTS5 for exact strings; OCR via PaddleOCR; Rust engine, GTK 4 app.
- No LLM, no network after model download.
Performance (i5 laptop with Iris Xe): right file first on 95% of a small test set, 4ms keyword results, 0.17s semantic search, Vulkan indexing 3x faster than CPU.
Missing: audio transcription, packages, Wayland shortcut on older desktops. Inspired by DigUp, a macOS app by Abdur Rahim.
More from Infra
- Virginia animal shelter says noise from nearby AI data center is terrifying its rescue dogs — Polymarket · 2026-10-11
- Virginia animal shelter says nearby data center noise is terrifying its rescue dogs — Polymarket · 2026-10-11
- Running Qwen3.8-27B on an RTX 5090 with vLLM: NVFP4 works, FP8 doesn't — BitGreen1270 · 2026-10-11
- Is AI Spreading Like a Virus? Reddit Essay Questions the $700B Datacenter Buildout — thepotatobake · 2026-10-11
- Running 426GB MXFP4 DeepSeek on 192GB VRAM to power a 4-sub-agent code review — HankYeomans · 2026-10-11
- awesome-local-ai: a task-sorted local AI tool list built for coding agents — blaizedsouza · 2026-10-11