Relations

Relations record how two images relate: same picture in two files, same subject drawn twice, an edit and its original. monbooru tracks four declared relation types, plus a rejection list:

Declare a relation by hand with Add a relation on any detail page (if the pair already carries one, Overwrite existing relation replaces it). The dialog spells the types out in plain words: “Same image (Duplicate)”, “Variant (Alternate)”, “Newer revision (Version)”, “Based on (Derivative)”. Related images render in their own panel on the detail page. When monloader pushes a booru post that declares a parent post, the pair is linked automatically as a derivative once both are in the gallery.

Collections are a separate, lighter grouping mechanism outside the relation graph.

Finding candidate pairs

monbooru computes a perceptual hash (a fingerprint of what the image looks like, one that survives resizing and re-encoding) for every image and uses it to suggest near-duplicate pairs. The queue of suggestions fills from several triggers:

• Automatically on ingest (the default; turn off with relations.incremental_on_ingest = false in the config file): each new image is probed against the index as it arrives, so fresh candidates appear on their own.
• Relations -> Find new pairs scans images added since the last scan.
• Settings -> Schedule -> Find relation pairs runs a nightly pass (off by default).
• Settings -> Maintenance -> Rebuild pair queue wipes the queue, including skipped pairs, and rescans from scratch.
• Relations -> Reset skipped returns previously skipped pairs to the queue without a rebuild.

How different two images may be and still get paired is set by Settings -> Relations -> Find-pairs default distance (default 4, range 0..12); lower means fewer, more confident pairs. If older images predate perceptual hashing, run Settings -> Maintenance -> Compute perceptual hashes once to backfill.

Pairs whose two images share a collection are left out of the queue by default - the collection already relates them. A per-collection find relations switch on the Collections page opts a collection back in.

The swipe session

Relations -> Start a session walks the queue pair by pair. Each pair renders side by side with the Hamming distance (how close the fingerprints are) and the larger file on the left (bigger usually means more canonical). Decide the pair (duplicate, alternate, version, derivative, not related) or skip it; deciding also drops the pair’s co-grouped rows so the session moves straight on. A metadata table compares the pair underneath; W swaps the two sides and C opens a compare slider.

Session order is set under Settings -> Relations: smallest distance first (default, most confident first), largest file first (handy when merging into the best-quality original), or random.

Browsing what you declared

Relations -> Browse relations lists every declared relation, one tab per type - Same image, Variants, Revision history, Based on, and Not related (so a rejected pair can be found and undone). Rows sort by most recent, size, or when the original was added. From here you can unlink an edge, dissolve a group, and merge same-image groups by selecting several and merging them into one.

Cleaning up duplicates

Two cleanup tools live under Relations -> Duplicates, one per kind of duplicate:

• Duplicate images walks the declared duplicate groups. Each row shows the original and one other member; Delete removes the non-original from the gallery, and Copy tags (a separate step) previews and layers the duplicate’s tags onto the original first. Delete all duplicate images removes every non-original member of every group in one go.
• Duplicate files walks byte-identical files stored at more than one path on disk - one image in the library, several copies in the folder. Delete removes the extra path and its file, keeping the canonical one; Delete all duplicate files runs it across every alias.

They are deliberately separate: a SHA-256 duplicate is the same file twice on disk (already a single image in monbooru), while a marked duplicate is two distinct images you declared to be the same picture.