Calibre-Web Automated

Last reviewed Sep 13, 2026 · 789 words

Drop an EPUB, MOBI, AZW3 or PDF into one folder and about a minute later it appears in your library, converted to your preferred format, with metadata fetched and the cover set. That is Calibre-Web Automated, and it is the one feature the original Calibre-Web always lacked. The catch to know before your first upload: the ingest folder consumes what you put in it. Files are moved into the library structure and the originals are gone from the drop point, so never point it at the only copy of anything.

What it adds on top of Calibre-Web, precisely

Calibre-Web is a browser front end for a Calibre library database; it reads and serves books but has always been awkward at adding them, because that job belonged to the desktop Calibre app. Calibre-Web Automated bundles the Calibre binaries inside the container and wires three things around them: a watched ingest directory, automatic conversion on import with EPUB as the default target, and metadata enforcement that writes changes you make in the web UI back into the book files themselves. Everything else, the reader, the OPDS feed, Kobo sync, send-to-Kindle by email, user accounts and shelves, is the upstream feature set, inherited and unchanged. It is a fork with a purpose, not a rewrite, and the project has been around since 2024.

Setup is three volumes and one default password to change

services:
  calibre-web-automated:
    image: crocodilestick/calibre-web-automated:latest
    ports:
      - "8083:8083"
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Europe/London
    volumes:
      - ./config:/config
      - ./ingest:/cwa-book-ingest
      - ./library:/calibre-library
    restart: unless-stopped

The first login is the upstream default, admin with password admin123, and changing it is the first click. The /calibre-library volume must contain a metadata.db; the image creates an empty one if the folder is bare, or you can copy in an existing Calibre library and it is picked up as-is. A 512 MB RAM floor is honest for browsing; conversions spike higher for a few seconds, and a big PDF-to-EPUB job on a Raspberry Pi is slow but completes.

The three ways people break it

Running desktop Calibre against the same library at the same time. Both write metadata.db and neither locks it for the other, so you end up with a corrupted database or vanished books. Pick one owner of the library folder; if you keep desktop Calibre, use it only with the container stopped.

Mounting the ingest folder from a network share. The watcher relies on filesystem events, which do not fire reliably over SMB or NFS, so books sit unprocessed. Use a local path on the host and copy files into it, or run the container on the machine the share lives on.

Dropping a folder of 800 books on day one. It works, but conversions are serial, and the queue can take hours with no obvious progress indicator beyond the container log. Batch in fifties for the first run and watch the log once.

Where the parent project still wins

If you already manage the library from desktop Calibre and only want a web reader, plain Calibre-Web is the same UI with fewer moving parts and no bundled conversion tooling. If your collection is comics or manga, Komga and Kavita handle CBZ and series metadata far better than anything Calibre-derived. And the limits people complain about in Calibre-Web are inherited intact: single library per instance, a metadata editor that lags the desktop app, and a dated interface that the fork has polished rather than replaced.

Backups are one SQLite file plus the folder

The library is the library/ volume, with metadata.db at its root and one folder per author. Back up the database with sqlite3 metadata.db ".backup /backups/metadata-$(date +%F).db" rather than a raw copy while the container is running, then rsync the book folders. The config/ volume holds user accounts and settings in a second SQLite file, app.db, and is small enough to include every time.

What I'd do

Run it on a mini PC or NAS with the ingest folder on local disk and the library on whatever you already back up. Keep desktop Calibre off that library entirely. Leave the default EPUB conversion on, because it is the format every reader app and every Kobo handles, and set up the email send-to-Kindle once so the ingest-then-send loop needs no computer. For anyone shopping for their first ebook server, the ebooks category lays out the alternatives, but a reader who mostly wants "put file here, read anywhere" should start with this one.