Line: Operate¶
Enterprise analog: Platform engineering / SRE.
Covers: Shared runtime health for self-hosted and related infra.
Tracking: Notion Workstreams database — view: Operate Work. Gatekeeper: Platform Engineer. See agent setup.
Editorial Publish is line-specific production; Operate owns shared host health. Writing deploy uses Editorial.Publisher with Operate standards (secrets, host).
Operate knowledge (hosts & tools)¶
| Doc | Purpose |
|---|---|
| Host inventory | host.konstant, host.vps-legacy (ai-sandbox), access aliases |
| Konstant operations | Dual-run, bootstrap, runbook map, tunnel |
| VPS legacy | ai-sandbox inventory, no-new-apps, migrate |
| Cutover and retire | Hostname flips, rollback, decommission |
| Operate tooling | secret-gate, DNS/tunnel CLIs, platform-control, pipeline scripts |
| Heritage infrastructure | Incubator / infrastructure-management decisions (non-binding) |
Secrets: always secret-gate → Bitwarden. See secrets standard.
Operate vs Software¶
| Concern | Owner |
|---|---|
| Application features, most product bugs, promote, app-repository deploy wiring | Engineering Lead (Software workstream) |
| Shared hosts, sockets, nginx/certs patterns, observability, incidents, retire | Platform Engineer (Operate workstream) |
| First-deploy shared platform and cert wiring | Platform Engineer |
| First-deploy app-repository deploy wiring | Engineering Lead |
First deploy may run concurrent Software and Operate workstreams on the same Initiative. Collaboration is allowed; accountability is not shared. Do not label ownership Joint.
New product features belong in a Software workstream. An observation may move the Operate workstream to Change or Incident, or add a Software workstream for an application defect. Do not absorb Software responsibilities into Operate.
Workflow¶
┌─ Change ──┐
Run → Observe ──┤ ├──→ Run
└─ Incident ┘
│
└──→ Retire
| Stage | Inventories |
|---|---|
| Run | ops-profile (pairs with Software host×shape) |
| Observe | observability depth |
| Change | change-style.hotfix \| normal \| migrate-host |
| Incident | severity → runbook depth |
| Retire | archive / migrate / delete |
Change and Incident are branches from an active runtime (Run/Observe), not steps in a mandatory sequential ladder. After Change or Incident work completes, the workstream returns to Run unless Retire is the exit. The exact entry criteria, exit evidence, and return-to-Run conditions must be defined before Conductor may authorize those transitions deterministically.
Host migration (VPS → konstant) uses Change with migrate-host then Retire on the legacy host — see cutover.
Roles¶
role.platform-engineer, role.on-call
Cross-line coordination¶
- In: Software deployment adds or activates an Operate workstream at Run.
- Out to Software: If Observe or Incident finds an application defect, add or activate a Software workstream on the same Initiative.
- Evidence: Major Change, Incident, and Retire outcomes are recorded on the Operate workstream and Initiative decision log.
Tools¶
Canonical catalog: Operate tooling.
Summary: konstant/VPS runbooks (in Conductor), systemd/nginx, MCP SSH, deploy workflows, secret-gate, cloudflare-tunnel-cli, hetzner-dns-cli, platform-control.
Tooling: Runtime artifacts and state live in runbooks, hosts, observability, and SSH-accessed systems. Notion owns the Initiative and Operate workstream state and links to operational evidence. Repository-local GitHub issues are optional for application defects. See tooling-map.md.