UNI Universal Natural Intelligence

Wiki · The Broadcast Platform

Receipt — Phase XI: observability heartbeat deployed + first row proven

The Broadcast Platform · production/docs/receipts/heartbeat_deploy_2026-07-12.md @ 44baf03d5041 (gen2-runtime) — opens the published snapshot ac338733bbba

How to read this page

A Plain and a Clear version of this page have not been written yet. What follows is the document itself.

Precise — the source document

This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.

Status: heartbeat deployed, enabled (boot-persistent), and writing valid ledger rows. Full alert-fire closure (G-OBS → corroborated) still needs an operator-supplied UNI_NOTIFY_URL.

Node uni-lab-79740c. Installed production/scripts/{heartbeat,notify,colony_gate}.sh → /opt/uni/production/scripts/ (executable), uni-heartbeat.{service,timer} → /etc/systemd/system/, systemctl enable --now uni-heartbeat.timer (symlink in timers.target.wants — fires every 60s + OnBootSec=60s). All via the uni-lab MCP, approval-gated.

First row (captured — /var/lib/uni/broadcast/audit/heartbeat.ndjson)

{"ts":"2026-07-12T00:27:25Z","p1_gate_pass":true,"p1_gate_rc":0,"relay_program_ready":"false",
 "colony_reachable":false,"colony_summary":"","colony_gate_pass":false,"colony_gate_rc":127,"fails":2}
  • p1_gate_pass: true — the load-bearing signal works: the heartbeat re-ran verify_p1.sh (throwaway container, exit 0) and confirmed the platform is genuinely up — not from process existence (CLAUDE.md method rule 1). A future platform regression will flip this to false.
  • relay_program_ready: "false" — honest idle state (nothing broadcasting; the relay has no program). Correct per docs/STUDIO_SYSTEMS.md. Counted in fails but it is an expected-idle condition, not a platform fault.
  • colony_reachable: false, colony_gate_rc: 127 — colony source is down (forage RED WITHHELD), and verify_colony.cjs/node are not present on the broadcast node (rc 127). Honest degradation.
  • notify path wired: notify.sh no-ops without UNI_NOTIFY_URL but logged exactly what it would have sent (/run/uni-heartbeat.err.log): [heartbeat] 2 check(s) failed ... (p1_pass=true relay_ready=false colony_gate_pass=false). The system python3 is present and parsed the relay JSON (no jq/venv dependency needed).

To fully close G-OBS (operator + one small follow-up)

  1. UNI_NOTIFY_URL in /etc/uni/runtime.env (a webhook the operator watches). Then the plan's Phase XI gate can be captured: systemctl stop uni-bcast-relay → within 60s the next tick logs a real failure and POSTs the webhook → systemctl start uni-bcast-relay → next tick clears.
  2. Idle-vs-live cadence / relay-not-ready semantics (known TODO in heartbeat.sh): during idle, relay_program_ready=false is expected, so it should not count as a fault when no broadcast is active. Until refined, idle rows honestly read fails:2; this is design noise, not a platform alarm (p1_gate_pass is the true health signal).
  3. colony_gate on-node: ship viewer/verify_colony.cjs + a node runtime, or point the heartbeat at the colony /producer/health over WireGuard, so the colony-size gate can actually run here.

The durable artifact is the committed production/scripts/heartbeat.sh + the running timer; this receipt is its first captured row.

sha256 b5dc78cb6e0fe1d9 — of the original file, so what was ingested stays checkable.