Data bundle 2026-11-04

Data bundle for restoring a BoothDex instance without the initial crawl: a snapshot from 2026-11-04 with 217,494 items (206,759 listed) and the archived images of delisted items.

File	Contents
catalogue.dump	PostgreSQL dump (custom format): items, BOOTH image links, taxonomy, schema, semantic-search vectors. No user accounts, sessions, favourites, viewed history, saved searches, job history, search texts or AI settings
images.tar	Archived images of delisted items (images/<sha1>.webp)
SHA256SUMS	Checksums

• Item vectors were made with Qwen/Qwen3-Embedding-4B (1024 dimensions). With another embeddings model, clear them once: update items set embedding = null.
• Job schedules are the defaults: hourly incremental crawl, enrich and embed, daily full sweep (03:00 JST); refresh is off. Current images turn on the daily details pass (fetches missing galleries) at startup.

See Deploying for the restore steps. The hourly crawl adds items published after the snapshot.