How I Turned Joanna Into an AGENT Engineer for Home Assistant Infrastructure Issues

n

n

I have been leaning harder into a pattern I like: let Home Assistant detect the problem, then hand the investigation to Joanna as an AGENT engineer with enough context to actually work the issue. That is exactly what I built out in this round of changes.

The point was not to create one more notification. The point was to create a reusable dispatch layer that can deal with all sorts of infrastructure issues once a package knows something is wrong.

What changed

I added a new Synology DSM monitoring package that converts native NAS entities into stable problem sensors, opens Repairs issues when the problem is sustained, and dispatches Joanna with the exact entities and diagnostics she needs.

I also updated processmonitor.yaml so the root filesystem can escalate in two stages. At 80 percent it asks Joanna to review disk growth before the box gets into trouble. At 90 percent it opens a Repairs issue and hands off an urgent dispatch.

The common thread is the shared joanna_dispatch helper. That script is the contract between detection inside Home Assistant and the heavier remediation logic that lives outside it.

Why I like this pattern

• Home Assistant stays good at sensing, timing, and routing.
• Joanna gets a scoped request, entity IDs, diagnostics, and guardrails.
• I can reuse the same dispatch contract across NAS health, disk pressure, network checks, container outages, and whatever comes next.

That is the piece I wanted to emphasize. Joanna is not just a notification sink here. She is the AGENT engineer handoff. Home Assistant decides when the condition is real, then passes enough structured context for the next layer to investigate infrastructure problems instead of just shouting about them.

The Home Assistant side stays small

The actual package code stays compact because the intelligence is in the handoff. Here is the short snippet from the process monitor package:

- service: script.joanna_dispatch
  data:
    source: "home_assistant_automation.disk_use_alarm"
    summary: "Home Assistant root disk usage exceeded 90%"

That is enough to turn a disk pressure condition into a real engineering request instead of another dead-end alert. The same model now works for Synology monitoring too.

Result

I now have Home Assistant packages that do more than detect trouble. They create a clean bridge into Joanna so infrastructure issues can be investigated with context, not guesswork. That is a much better place to be when the goal is to keep a homelab stable without turning every warning into manual babysitting.

I am still using the same kind of webhook-first pattern I have trusted for years, and I am logging the handoffs so they are easy to trace later in the Home Assistant logbook.

Draft generated with help from Codex