Downloading

Sending URLs to the queue

The home screen is a single command bar. Paste a URL and press Enter; the job is queued and you land on the queue screen to follow it. Anything a gallery-dl extractor matches works: a single booru post, a pool, a tag search, an artist page, a manga or comic gallery, or a direct link to an image file.

Two other ways in:

• The monsender extension sends the URL of the page you are browsing with one click.
• Scripts can POST /api/v1/queue with a bearer token; see API and development.

The add bar also accepts a bare md5 hash to import the booru post carrying that file; see tagging by hash below.

What happens to a job

A queued URL is a job. The worker first resolves it with gallery-dl to learn which posts it expands to - one job becomes one or more items. Each item is then downloaded to scratch space, its metadata is mapped onto monbooru’s model, and the file is pushed into the target monbooru gallery over the API. Files are deleted from scratch space once monbooru has accepted them; monloader keeps no copy of anything.

Per-item outcomes

Every item ends with one of six outcomes, never silently dropped:

monbooru ingests jpg, png, webp, gif, mp4, webm, and cbz files; a download of any other type (audio, svg, a document) is skipped_unsupported rather than pushed and rejected. A failed item carries a stable machine-readable error code; the full table is in API and development.

Each job row aggregates its items into summary counts (created / duplicate / enriched / skipped / failed / canceled), so a single post that was already saved reads as “duplicate” at a glance.

Large searches: the cap

A tag search or an artist page can expand to thousands of posts, so a job takes at most max_items_per_job posts (default 200). The cap is never silently applied: a capped row notes that more is available and offers two actions:

• get next N fetches the next window of the search (N being the cap).
• get all keeps fetching window after window until the search runs short.

A search and its continuations collapse into one queue row with summed counts. Booru pools and manga galleries are exempt from the cap - they are one work you asked for as a unit and always come down whole.

Retry, force, and cancel

• retry re-runs a job. For a bulk search, posts already fetched are skipped (skipped_archive). Not shown on a job where everything succeeded.
• force download appears when a job has skipped items: it bypasses the skip and fetches them again - useful to re-import a post whose image you deleted in monbooru.
• Re-submitting a single post always re-fetches it: monbooru recognizes the file and merges any new tags into the existing image, so this is how you refresh one post from its source.
• cancel stops a running job; items still in flight end as canceled, finished ones keep their outcome. remove clears a finished row (for a search series, the whole series), and clear next to the add bar removes all finished rows at once (it asks first).

Finished rows also go on their own after a week, so the queue does not fill up with downloads you have long stopped caring about. The status column shows how long each row has been in its state (“5m ago”, “3d ago”), with the exact time on hover, so you can see what is about to go. Change the week in Settings -> downloads -> clear history after (days), or set it to 0 to keep rows until the queue’s own limit of 100 pushes them out.

Pause and resume

The pause button in the topbar of the queue, PTR and settings pages (and in the monsender popup) holds all downloads globally: the job in flight finishes, then nothing new starts until you resume. Use it to queue up a batch before letting it run.

The queue lives in memory: a restart clears pending jobs, the recent history, and the pause. That is fine in practice - re-submitting a URL is cheap, since already-fetched posts are skipped and already-pushed files land as duplicates.

Pools and manga

Multi-page works are handled by kind:

• A booru pool (an ordered set of posts forming one work) imports as an ordered collection: each page is pushed as its own image under a shared collection label (the pool name) and page number, keeping each page’s own tags.
• A manga or comic gallery is bundled, in order, into a single .cbz file (a comic book archive) and pushed as one file, which monbooru opens in its reader. The bundle’s tags are the union across the pages and its rating the strictest seen. If any page fails to download, the job fails rather than pushing a truncated book.
• A manga title page that lists chapters imports each chapter as its own .cbz.

Some URLs list other pages rather than files - a forum thread whose images are hosted elsewhere, an archive board. monloader follows these handoffs and imports the files they lead to as loose items, bounded by the per-job cap.

Tagging by hash

Two features find a post from a file hash instead of a URL; both walk the same lookup chain:

• Hash import: paste a file’s md5 (bare or as md5:<hash>) into the add bar and the matching booru post is imported like any single post.
• Lookup enrich: monbooru’s “fetch tags” button asks monloader to find tags for an image it already holds and merge them in, without re-downloading the file.

Both show up on the queue page as jobs like any other.