Repository navigation
v2.5.1
HexDB 2.5.1 reads the text in scanned documents and pictures, makes thumbnails of many more kinds of files, and adds EF Core 10.
New
- OCR. With Tesseract installed (the Docker image includes it, with English), HexDB reads the text of images and of PDFs with little text of their own (scans: their first
files.ocr_pagespages, 20). It goes inmeta.text, so$textsearches find it, withmeta.ocrsaying how it was read. Text that doesn't read like words isn't kept, so photos don't fill the index.files.ocr_languagespicks the languages (eng;eng+deu, ...). - More thumbnails. SVG drawings, a PDF's first page and RAW camera files (the preview the camera stored) get thumbnails out of the box. With ffmpeg installed (the Docker image carries a small build of it), so do HEIC and AVIF photos (phones' tiled HEIC included), HEVC, VP9 and AV1 video, and WebM, Matroska, AVI, MPEG-TS and similar files.
- Outside tools stay contained. Tesseract and ffmpeg (
files.tesseract,files.ffmpeg:auto, a path, oroff) never see a file on disk: they read from a pipe and run in the process sandbox, with no network and nothing to write./statuslists them understorage.files.tools. - EF Core 10.
HexDB.EntityFrameworkCoresupports EF Core 10 on .NET 10, and still EF Core 9 on .NET 8, from one package;HexDB.Clienttargets both. - VS Code extension 0.3.0 shows the delete history, with restore and purge.
Upgrading
Restart the hexes one at a time, as usual: replicas first, the Overseer last. Every file's metadata is read again in the background (extraction version 4), which adds OCR text and the new thumbnails to files already stored; on a large store this takes a while.
The Docker image is now based on Debian trixie (it was bookworm), for Tesseract 5.5 and ffmpeg 7.1. HexDB's own binaries are unchanged.
Full Changelog: v2.5.0...v2.5.1