FreeOSBot is an autonomous AI DevOps engineer running continuously on freeos. It manages real infrastructure, pushes real fixes through GitOps, watches the cluster every 30 minutes, runs 8 deterministic checks, and leaves an audit trail. No theatre. No fake dashboards. Just work.
FreeOSBot is not "AI for DevOps" as a slide. It is the operator for a real HealthCloud stack. It diagnoses, patches, commits, deploys through GitOps, verifies outcomes, and keeps memory of what happened.
FreeOSBot is the master brain. Watchmen are the narrow sensors. Sensors observe; the master reasons and acts. Complexity stays central. Monitoring stays simple.
Cluster changes are committed and deployed through ArgoCD. No cowboy shelling into the cluster. If it is not in git, it is not state.
Traefik, Longhorn, Velero, Kafka/Strimzi, Mirth Connect, Keycloak, Vault, Wazuh, Prometheus and Grafana are all part of the actual operating surface — not decorative copy.
Session exports, daily memory, learnings, error logs, and Git history mean this system can reconstruct context and explain what it did.
Not claims. Work completed on the live stack today. This is what competence looks like when you stop writing brochure copy and start keeping score.
Diagnosed the compaction loop, moved the system to GPT-5.4 with 1M context, and removed the underlying failure mode instead of just damping symptoms.
Resolved VM memory ballooning across 7 KVM VMs and recovered 28GB live, without downtime.
Created 8GB emergency swap and made it persistent in /etc/fstab so the host has breathing room under pressure.
Repaired iframe issues, broken images caused by private repo auth, missing PDFs, committed the fix, and deployed it via GitOps.
Found HTTPS redirect logic blocking ACME challenges, patched the Traefik IngressRoute, and restored correct certificate issuance.
Node health, pod failures, ArgoCD sync, cert expiry, disk pressure, Velero backup, Watchman heartbeat, and self-health are now all checked on deterministic cron.
Connected NFSv4 backup to QNAS and confirmed first k3s SQLite and system config backups landed off-host.
Created a 25-category DevOps checklist with 150+ checks for serious multi-cluster operations.
The pattern is simple: FreeOSBot is the master brain, Watchmen are the per-cluster sensors, and GitOps is the control path. That keeps reasoning central, actions auditable, and the cluster stable.
This is today's operation log rendered as site copy. It is not a staged stream and not a toy dashboard. Just the work that mattered.
Identified the compaction loop, switched the platform to GPT-5.4 as primary with 1M context, and removed the structural cause instead of papering over the symptom.
Resolved memory ballooning without downtime. Live recovery, not maintenance-window theatre.
Created an 8GB swap safety net and persisted it in /etc/fstab to reduce host fragility under pressure.
Repaired iframe routing, broken images caused by private repo auth, and missing PDFs. Committed and deployed through GitOps.
Tracked ACME challenge failure to an HTTPS redirect issue in Traefik and patched the IngressRoute correctly.
8 cron-based checks now cover node health, failing pods, ArgoCD sync, cert expiry, disk pressure, Velero, Watchman heartbeat, and self-health.
Wired NFSv4 backup to QNAS and confirmed k3s SQLite and system config backups landed off-host.
Produced a 25-category, 150+ check DevOps checklist for serious multi-cluster operation and review.
FreeOSBot and the Shadow Architecture exist to make sovereign Health IT commercially viable for hospitals, regions, and hosting partners — with predictable spend and lower operational drag.
Designed to remove hyperscaler tax, VMware-style licensing overhead, and avoidable operational waste from Health IT platforms.
Run on-prem in your own environment, or through a trusted hosting partner. Your jurisdiction, your governance, your infrastructure choices.
Simple recurring pricing. No hyperscaler-style rent, no egress surprises, and no exploding invoice because usage drifted during the month.
12 months of stewardship included. Then a fair subscription while we continue improving the platform. Pay it out anytime and take full ownership. This is how infrastructure software should work.
If a system claims competence, it should expose the work: architecture, manifests, reports, and the state of what is live versus what is still under construction.
Shadow Architecture, operating model, failover thinking, and the separation between sensors, reasoning, and GitOps control.
Open PDF →Component inventory, deployment assumptions, and technical boundaries for the FreeOSBot stack.
Open PDF →What this replaces, why it matters, and why sovereign infrastructure should not need hyperscaler rent.
Open PDF →Live, read-only cluster metrics, updated every 15 seconds. Real data from the underlying k3s cluster.
Open Dashboard →FreeOSBot is built and operated under eHealthBrains. Danish-built, healthcare-native, and not interested in sounding like Silicon Valley.