The installer pointed it at the node's own panel and left AGENT_METRICS_PORT
unset, so the agent refused /prom/… and core never saw it. A node's own panel
now ignores the local Prometheus and shows no install buttons.
The status handler probed Prometheus before reading the job, so a poll
could get exit 0 with ready=false; the page then stopped polling and the
Monitor tab waited for App's 60 s check.
- UI: settings tab with sections (server, API keys, Prometheus, audit, migrate,
about); server settings moved there; master keys table rework (menu, chips,
version, compact phone cards); three-state node version chip
- I1: random per-server template (DNS / QUIC Initial / STUN) kept in users.db,
GET/POST /api/masking, change from the master keys banner
- Hub rejects nodes older than 0.5.0 (close 4026), /ui/nodes reports outdated
- install.sh: spinner for long commands (full output in the log), step progress
bar, cards; installs older than 0.5.0 require a full reinstall with data reset
- Bump to Beta 0.5.0