The probe fired the instant a batch finished, but a fresh drop takes ~1-2 min to appear even with the server's auto-scan ON — so it always missed and we crawled anyway, defeating the optimization. Now probe_present_libraries POLLS each candidate over a grace window (probe_grace_minutes, default 2), skipping a library's crawl as soon as the server reports it has the item, and only crawling what's still missing when grace expires. The probe target for a media type you DIDN'T just download is an old item the server already has → confirmed instantly, no wait. grace=0 probes once. |
||
|---|---|---|
| .. | ||
| handlers | ||
| __init__.py | ||
| api.py | ||
| blocks.py | ||
| deps.py | ||
| progress.py | ||
| schedule.py | ||
| signals.py | ||