Long-term items, roughly in the order they're actually likely to get picked up. See the project's own claude-md/roadmap.md for full status detail — this page is a lighter summary.
- Fleet automation / self-healing ("more JARVIS, less manual sweep") — nightly deterministic health sweeps, hourly Discovery pushes, and a weekly LLM-driven judgment review are built and live. This is largely done; the remaining piece is extending fleet-wide compose-drift detection, which is also now built.
- PBS backup integration — a local-hardware Proxmox Backup Server being wired in as a backup target for hot-bm-nl's VMs. Real open question: the local network's bandwidth may not be sufficient for full VM image backups at this data scale — under active investigation.
- Edge load balancing — a second edge VPS or per-domain failover for hot-pn/hot-erp-nl, which currently have no public path independent of the Gateway. Three options scoped, none chosen yet.
- HoT Sync (Flutter) — personal-services consolidation (Nextcloud/Immich/Notesnook/Vaultwarden). Needs a landing spot now that sn-personal is permanently retired.
- HoT Command (Flutter) — mobile ops dashboard.
- Second bare-metal node for real HA.
- Terraform/Ansible IaC — the fleet is currently hand-configured +
hot-config git tracking, not declarative IaC.
- hot-erp hosting client companies' backend software — stated future goal, no timeline.
- sn-personal (VLAN 40) and pn-test (VLAN 60) — both retired outright, Mr. Byrne's decision. Any future personal-services work needs a fresh landing spot, not a revival of these.