Docker Hub data-image popularity intelligence (monthly)

Popularity intelligence for 75 curated data/ML-ecosystem container images on Docker Hub (September 2026 snapshot, from the keyless official Docker Hub v2 repositories API). Each row is one image (primary key: repo path namespace/name) with its curated category (15 classes: databases-relational, databases-nosql, cache-kv, search, timeseries, streaming, etl-orchestration, notebooks, ml-platform, vector-db, monitoring, bi, db-tools, object-store, columnar-analytics), lifetime pull_count, star_count, last_updated, days since the last tag push, a documented 0-100 popularity_score = 100*(0.50*min-max log10(pulls) + 0.30*min-max log10(stars) + 0.20*recency) ranked as popularity_rank (ties: pull_count desc, image asc), an official-image flag, a short description, and a per-row hub.docker.com link. Entity resolution: repo path is the identity; Docker Official Images resolve to library/<name>. Pull counts are cumulative and include CI reinstalls/mirrors (distribution volume, not unique users); they tick continuously, so same-month live re-runs yield new snapshots. Agent-curated by intel-1 (2026-09-25T22:05Z): collection = 75 polite keyless API calls (0.5s pacing), transformation = curated taxonomy + popularity scoring. Pull/star counts and update timestamps are non-copyrightable facts — commercial_use = yes, with Docker Hub attribution. Sample use: order by popularity_rank for the most-pulled data images, or filter category = 'vector-db' for the vector-database leaderboard.