DevOps

Experimental · mode-gated actuation An SRE agent whose proposed infrastructure changes must pass checks outside the model. I set the architecture, governance rules and product direction, direct AI-assisted implementation, and operate the test environment. What works—and the limitsThe public repository reports deployed triage, an operator console and mode-gated actuation in an owner-set Semi-auto mode. Missing or invalid mode state falls back to read-only Shadow. The product landing page still says mutation is off; the newer public README reports it enabled. This review follows the README for implementation status, not a fresh inspection of runtime settings. Operational assurance is still being evaluated. ...

Agentic ChatOps triages infrastructure alerts and predicts the consequences of each proposed fix before it acts — auto-resolving only the remediations a causal world-model can prove safe, behind a fail-closed, machine-verified prediction gate. GitHub: papadopouloskyriakos/agentic-chatops 390 / 1149causal world-model nodes / edges 71active n8n workflows 58SQLite tables · 30 versioned 6,186 · 12% autogoverned decisions 12 / 14agent-guide rubric dims at A 1,561autonomous agent sessions 792alerts auto-resolved 4.10 / 5avg judge score · 1,511 sessions Usage The auto-resolve rate is intentionally conservative. Autonomy is earned per-action-class through the verification contract — an action type may self-resolve only after its predictions have repeatedly matched observed reality — never switched on globally. A low number here is the safety system working, not a ceiling. ...

Everything in the homelab — network devices, VMs, Kubernetes workloads, Docker services, backups — is managed through GitLab CI/CD, Atlantis, and Argo CD. The aim is to make intended configuration reviewable and repeatable; manual recovery remains an operational concern. I own the architecture, integration and operation of this lab. Separating platform provisioning from application delivery makes changes easier to review and repeat. It does not make recovery automatic; configuration drift, provider failures and restoration still need checking. The next step is continued reconciliation and recovery validation, not a customer launch. ...