24/7 Autonomous Operation — Enhancement Plan
24/7 Autonomous Operation — Enhancement Plan
Current state
Fleet healthy: 8 agents up, memory restored, A2A working. But "always-on autonomous" needs hardening beyond "works right now."
What to enhance (prioritized)
1. Supervisor / self-heal loop
Agents restart on crash (Restart=on-failure), but nothing watches for hung lanes (low CPU + message/send timeout = blocked provider call, not a dead process). Add a watchdog that detects turn-stall and restarts a stuck agent — the single highest-leverage fix for true 24/7.
2. Unified memory bus
Eve (TS) and Tam/Eva (laptop) sit outside the solos.db memory plane. Route everyone through one durable store so context survives restarts and is shared across agents. This is the "brain" living up to its name.
3. Single cron dispatch surface
Add laptop agents (tam/eva) to central-cron AGENT_PORTS; retire per-profile Hermes cron. One scheduler, one delivery target per job, read-back verification — silent failures are unacceptable.
4. Provider failover + cost ceilings
CostGov is the budget gate; ensure every lane has a fallback chain + 429/401 handling so a provider flap cascades down, not dead-ends. Two workspaces already exhausted provider capacity in one session — meter and cap, don't burn lanes on parallel fan-out when one lane suffices.
5. Health/drift telemetry
Verified ground truth over self-reports: sha pull-back, read-back, /health vs live message/send. Add a periodic fleet self-audit that posts proof-of-life even when green (the "silent-when-green" job is the visibility proof).
6. Secrets consolidation
Move inline systemd env (Eve's keys) into the central config vault (0600), never in unit files. Reference by path only.
7. Port/path single source of truth
The 8432 vs 8433 drift cost real debugging time. Generate the port map + paths from one config, not from scattered notes.
Bottom line
The brain/muscle split is correct; the gap is not architecture but robustness — supervision, unified memory, single scheduler, failover, telemetry, secrets. These seven items are what make the fleet genuinely 24/7 and autonomous.
SolOS Architecture Review — Verification Findings (Tam)
Evidence-grounded review. Every claim below was checked against live system state (curl health, port scans, systemd units, and actual source under /home/ubuntu/.solos/ read via read-only tam_server_key). No claims from memory.
Blueprint v2 — Final Audit (Tam, research lane)
Directive: Young → Atlas → Tam. Verify tunnel, audit blueprint vs live, list blockers, register. Read-only posture; nothing modified on server.

