SUCCESS_COUNTS.md

Successful-run counts

The shared backend records Object and Scene generation separately, only after their GPU calls return successfully. The Scene Space embeds the same backend, so a scene generated through either entry point counts once for Scene. Mesh extraction, camera previews, downloads, visits, and failed requests do not count.

Badges at the right edge of each title row show the last 30 days' total and today's successful runs. The Daily success history panel contains the daily table for today and the preceding 29 days. Dates use America/Los_Angeles. History before instrumentation starts is unavailable; the total includes the available recorded activity without waiting for a full month.

The private dataset CaiYuanhao/space-success-stats stores the aggregate counts. Add a long-lived fine-grained token with read/write permission only for that dataset as the SPACE_STATS_TOKEN Secret on DiffusionGS-Object. The static Scene Space requires no secret. SPACE_STATS_REPO can optionally select another initialized private dataset. Never put tokens in repository files. The expiring interactive OAuth login is not used as a runtime credential.

For a differently named Secret, set the non-sensitive SPACE_STATS_TOKEN_NAME Variable to that name. Existing deployments use cyh_track_liuliang.

Recording runs in the CPU caller after the ZeroGPU worker returns, avoiding forked-process state or temporary-disk persistence. Optimistic commits and event IDs handle concurrent writers and uncertain commit responses without double counting. No images, prompts, IP addresses, or user IDs are stored. An outage does not prevent returning generated results; failed counter writes produce an anonymous SUCCESS_COUNT_NOT_SAVED log entry for recovery after retries.

Badges load on page opening and update automatically after successful generation. Open Daily success history and click Refresh counts to reload from storage. Cached remote reads expire after 30 seconds.

python -m unittest discover -s tests -p test_space_success_stats.py -v