LANrurugi
Self-hosted manga & e-book reader
A self-hosted library for manga first, and for e-books such as PDF and EPUB too. Built on LANraragi and extensively reworked and extended in Rust + React; your existing LANraragi library works as-is. It includes in-page manga translation, recommendations for similar works and AI plugin generation.
第1話
Translation, plugin writing and tankoubon sorting can be handed to AI. Every AI feature is optional, and everything else works without an API key.
Manga-specific OCR reads vertical and horizontal text, a glyph-level mask erases only the lettering, and the translation is set back into each bubble in the original direction. A glossary keeps names consistent across a volume, and a per-volume font cache keeps the lettering consistent.
Describe the site and the AI writes a login, metadata or download plugin, test-runs it in a Deno sandbox and fixes it from the error if it fails. LANrurugi fetches every page itself, and test credentials are never sent to the AI.
A local ONNX embedding model shortlists candidates from your whole library, then an LLM ranks them. Tankoubons are matched on their last volume, so volume 1 is never recommended back.
A local model finds loose archives that belong to the same series and suggests grouping them into a tankoubon (no API key needed for this step); the AI then suggests the series name, chapter titles and reading order from the volume and chapter numbers in the filenames. A whole subfolder can also become a tankoubon in one step.
When no metadata plugin finds tags, an LLM fills them in from the title and existing information.
Describe the kind of works you want to follow and the AI turns it into the subscription's filter rules, which are validated before they are saved.
第2話
All volumes of a tankoubon form one continuous run of pages; page turns cross volume boundaries, and progress and bookmarks are kept per volume.
One-click presets for cover, color pages and afterword, number keys to mark chapters, and AI-generated chapter names shown first.
Bookmarks point at a page; stamps can mark a region of it, and placing a stamp bookmarks the page.
Thumbnails load on demand, so even an omnibus of several hundred pages opens quickly without the scrollbar jumping.
Formats the browser cannot display (such as AVIF, JPEG XL or animated images on some browsers) are converted first; large images can also be scaled down and sent as WebP or JPEG XL to save bandwidth.
Guests can browse only the categories you choose and cannot see anything else.
第3話
Downloads run on the server with real byte progress and time remaining, per-domain concurrency and rate limits, and no duplicate downloads of the same address. Works delivered page by page, such as on Pixiv, are bundled into one archive.
Subscribe by source site plus artist, tags, minimum rating, publication age and more; new works are checked on a schedule and downloaded automatically. A preview shows what will happen to each candidate, with one-click block and undo, and candidates are re-evaluated when their rating or tags change.
Uses the site's version chain to tell whether a new download is newer or older than the archive in your library. A newer one can replace the old archive, which is deleted only after the new one is imported; an older one can be refused, or you can be asked before it downloads.
Finds archives that hold several volumes, nested folders or archives inside archives, proposes a split and runs it in the background once you confirm. The AI names the resulting files, and tankoubon membership can be carried over and the original deleted.
When an upload largely overlaps an archive already in your library, the two are compared page by page and shown side by side. Drag the extra pages into place to create a page patch instead of replacing the whole archive.
Finds garbled (mojibake) names in archives, categories, tankoubons and tags, lists the repaired names for you to confirm, then applies them and updates the search index.
第4話
Keeps LANraragi's library layout and HTTP API, so third-party readers connect as before. Tags, categories and other metadata come over from a LANraragi backup; the two databases stay independent.
Settings → Database Backup/Restore → Backup Database, and save the JSON file. LANraragi can keep running.
Mount the same manga directory as LANraragi's library. Thumbnails and the database need mounts too, but in new directories, not LANraragi's.
kamimamita
Open http://<server>:3000 and sign in with the default password kamimamita (the same as LANraragi). The first-run wizard asks you to set your own.
Upload the JSON in the wizard's "Import from a LANraragi backup" step, or later under Settings → Database Backup/Restore. Tags, categories, tankoubons and stamps come back; reading progress is not part of the backup. Archives are matched by file name, so let the library scan finish first.
Search syntax
artist:"some artist" |
circle:example is:unread
size:>=100M -tag:sample
or / grouping / negation · read state · file size · progress ranges · date added · per-account search history
Supported formats
Archives ZIP / CBZ RAR / CBR 7Z / CB7 TAR / CBT LZH
E-books PDF EPUB MOBI / AZW3 DjVu FB2 TXT
Scanned pages are served as stored, never re-compressed. Text books are typeset into pages on the server, with a Chinese, Japanese or Korean font matched to the book's language, and the book's own table of contents becomes its chapter list.
最終話
Hit “Copy”, save it as compose.yaml, and change the directories as the comments describe.
docker compose up -d
On a NAS, create a compose project in Synology Container Manager, Portainer or similar and paste the same text.
http://<server>:3000
The first-run wizard walks you through replacing the default password, adding an API key for the AI features and importing a LANraragi backup.
# compose.yaml
services:
lanrurugi:
image: ghcr.io/thelastfantasy/lanrurugi:latest
container_name: lanrurugi
ports:
- "3000:3000"
volumes:
# Library — can point straight at LANraragi's content directory
- ./data/manga:/home/koyomi/lanrurugi/content
# Thumbnail cache, LANrurugi only — not LANraragi's thumb directory
- ./data/thumb:/home/koyomi/lanrurugi/thumb
# Redis data, LANrurugi only; LANraragi's metadata is imported from its backup JSON
- ./data/database:/home/koyomi/lanrurugi/database
# Embedding model for recommendations, downloaded on first use
- ./data/models:/home/koyomi/lanrurugi/models
environment:
# Set to true when served over HTTPS through a reverse proxy
- LANRURUGI_FORCE_SECURE_COOKIES=false
restart: unless-stoppedRunning behind Nginx, Caddy, Traefik or another reverse proxy? See reverse proxy examples
LANrurugi's thumbnail and database directories must be separate from LANraragi's — never the same path — and the two can run side by side. Tags, categories and other metadata come in through LANraragi's backup JSON, a step in the first-run wizard.
The same image is also on Docker Hub:
jade913/lanrurugi
Paste LANraragi's docker run command or compose.yaml; its library directory is put into the compose text above, thumbnails and the database stay in LANrurugi's own directories, and the port becomes the one after LANraragi's. The conversion runs in your browser only — nothing is uploaded.
Native packages are available for Linux (x86_64 / ARM64), Apple Silicon Macs and FreeBSD: install Redis, Deno and the other prerequisites with your package manager, unpack and run — or register it as a system service. On Windows, use Docker Desktop.
Native install guide →