Remove Performance tab, PerformanceStats component, settings toggle,
Web Vitals observers from tracking script, and all related API types
and SWR hooks. Duration tracking is preserved.
Performance metrics moved from dashboard into a new Performance tab.
Fixed null handling so "No data" shows instead of misleading zeros.
Script no longer sends INP=0 when no interaction occurred.
Peak Hours back in its original grid position next to Campaigns.
Search panel now placed below as a standalone row, and hides
entirely when there's no GSC data instead of showing zeros.
Dashboard: compact Search Performance panel showing top 5 queries,
clicks, impressions, and avg position alongside Campaigns.
Search tab: clicks/impressions trend chart, top query position
tracker cards, and new queries badge.
Search Console page with overview cards, top queries/pages tables,
and query↔page drill-down. Integrations tab in Settings for
connect/disconnect flow. New Search tab in site navigation.
All dates now use day-first ordering (14 Mar 2025) and 24-hour time
(14:30) via a single formatDate.ts module, replacing scattered inline
toLocaleDateString/toLocaleTimeString calls across 12 files.
Smooth out the jarring visual pop when loading skeletons are replaced
by real content. Only animates after an actual skeleton was shown —
cached data still renders instantly with no delay.
Replace manual useState/useEffect fetch patterns with SWR hooks so
cached data renders instantly on tab revisit. Skeleton loading now
only appears on the initial cold load, not every navigation.
New hooks: useFunnels, useUptimeStatus, useGoals, useReportSchedules,
useSubscription — all with background revalidation.
After refreshing the base token, call switch-context to get an
org-scoped token. This prevents 403 errors on Pulse API requests
when the access token is refreshed mid-session.
Remove the individual session journey page and make the live visitor
count a static indicator. Prepares for the new aggregated User Journeys
feature (v0.17).
Replaces 4 separate frustration API calls with single useBehavior hook.
Removes manual fetchData, loading/error state, and refresh interval—SWR
handles caching, revalidation, and error state automatically.
Frustration APIs and dashboard API are separate — when frustration
calls fail, Scroll Depth still rendered from cached SWR data,
creating a broken mixed state. Now tracks error state and hides
the bottom section entirely on failure.
- Remove column headers for cleaner look
- Show secondary info (avg, sessions, last seen) on hover
- Add orange percentage badge that slides in on hover
- Add empty row padding for consistent card height
- Add column headers to rage/dead click tables
- Rich empty states with icons matching dashboard pattern
- Add frustration trend comparison chart (current vs previous period)
- Show "New" badge instead of misleading "+100%" when previous period is 0
- Click-to-copy on CSS selectors with toast feedback
- Normalize min-height to 270px for consistent card sizing
- Fix page title to include site domain (Behavior · domain | Pulse)
- Add "last seen" column with relative timestamps
Add send hour, day of week/month selectors to report schedule modal.
Schedule cards now show descriptive delivery times like
"Every Monday at 9:00 AM (UTC)". Timezone picker moved into modal.
Add LRU cache provider (200 entries) to prevent unbounded SWR memory
growth. Clean up stale PKCE localStorage keys on app init. Cap chart
annotations to 20 visible reference lines with overflow indicator.
Replace 5-second setInterval polling with EventSource connection to the
new /realtime/stream SSE endpoint. The server pushes visitor updates
instead of each client independently polling. Auto-reconnects on
connection drops.
Replace useDashboardOverview, useDashboardPages, useDashboardLocations,
useDashboardDevices, useDashboardReferrers, useDashboardPerformance, and
useDashboardGoals with a single useDashboard hook that calls the existing
/dashboard batch endpoint. This endpoint runs all queries in parallel on
the backend and caches the result in Redis (30s TTL).
Reduces dashboard requests from 12 to 6 per refresh cycle (50% reduction).
At 1,000 concurrent users: ~6,000 req/min instead of 12,000.
