| Dataset Starts Why | |
| Spaces by type, SDK and linked models (spaces_by_*, space_creation_monthly) Mar 2022 Earliest Space record; only Spaces that still exist are counted | |
| Space visits (space_visits_monthly) Oct 30 2023 The visit counter on Space pages started then | |
| Repo creation, all types, including deleted repos (repo_creation_weekly_*) Nov 15 2021 First repo.create event | |
| ↳ split by client (browser vs huggingface_hub) Oct 19 2023 The User-Agent wasn't recorded on events before that (shown as no_user_agent_recorded) | |
| ↳ agent vs human ~Apr 1 2026 The agent/<name> tag only exists from then | |
| Hub request traffic, agent vs human (hub_requests_*_sampled) Aug 21 2026 The request logs only keep about 6 weeks. I took four 10-minute samples per day (166 samples, 42 days) | |
| Agent- vs human-created repos: survival and traffic (agent_vs_human_repo_cohort) Repos created Apr–Aug 2026 Matches when the agent tag started; each repo has had at least 30 days |