Counts approved keywords by processed_volume_trend_meta -> confidence (0-100), bucketed
in 10-point bands across all 200 shards × 256 ranges (~1-2 min). Examples are top-10 by
processed_keyword_volume per bucket. Leave country blank for the global view.
Per-cycle snapshots recorded after each run of keep_running.sh (post pe_update.py).
Stacked area = keyword counts per bucket over time; line below = mean confidence over time.
Each snapshot records 'global' plus the top-10 countries by total approved-keyword count
at that moment, so the country list may shift slightly across cycles.