Override NDK default outbox relays (purplepag.es DNS failure was stalling
getZapInfo), add 45s zap timeout, disable outbox on NWC instance, fix
follower notification dedup in poller. Zap history rows now show a
clickable preview of the original zapped note.
Threads now render the focused note immediately instead of showing a
loading skeleton. Root + ancestors fetch in parallel, timeouts cut
from 10s to 5s per round-trip, ancestor lookups from 5s to 2s.
Trending feed adds a base recency score so notes always appear even
when engagement data times out from relays.
Hidden dev tool showing NDK uptime, live subscription status,
per-relay connection state, per-tab last-updated timestamps,
and recent feed diagnostics log. Polls every 2s while visible.
Closes with Ctrl+Shift+D, Escape, or X button.
- Relay status badge in feed header: shows connected/total relay count with
color coding (green >75%, yellow 25-75%, red <25%), hover tooltip with
per-relay status
- Toast notification system: transient messages for connection lost,
reconnecting, relay reset, and back-online events
- Per-tab "last updated" relative timestamp in feed header (global,
following, trending tracked independently)
- Consolidated all relay management into RelaysView (removed duplicate
relay section from Settings); per-relay remove button on health cards
- Show all supported NIP badges on relay cards (was filtering to 11 notable)
- Tooltips on relay status dots explaining green/yellow/red/gray meaning
- Fix relay removal with trailing-slash URL normalization
- Fix stale health results lingering after relay removal
- Add acknowledgements section to README
Global feed now uses a persistent live subscription (closeOnEose: false)
so new notes stream in real-time instead of requiring manual refresh.
Inspired by Wisp's streaming architecture.
Every fetchEvents call across the entire codebase now uses
fetchWithTimeout with groupable: false — prevents NDK from
batching/reusing stale subscriptions. This fixes Articles, DMs,
Notifications, Zaps, and Trending hanging indefinitely.
Also adds since filters on global (2h) and follow (24h) feeds
to ensure relay freshness, and fixes ArticleFeed re-fetch bug
where follows changes wiped latest tab results.
The liveness probe in ensureConnected was causing a death spiral —
it force-disconnected working relays when a 3s probe timed out,
then resetNDK killed new connections before they could establish.
Now ensureConnected trusts relay.connected and only reconnects when
zero relays are connected. All fetchEvents calls have timeouts
(5-10s) so nothing hangs. resetNDK kept as background-only recovery
in the connection monitor after 30s of continuous failure.
Also adds feed diagnostics (feedDiagnostics.ts) for tracking fetch
timing, event freshness, and relay states via console helpers.
