Hi everyone,
Thanks again for the testing. The report showing that ZimaBrain counted alerts but did not explain them properly helped expose a real gap.
I pushed an update today with these fixes:
- Real Alerts now try to map back to the source, object, and condition where the dashboard evidence allows it.
- Info Only items are now listed instead of only showing the count.
- Unsupported / N/A metrics are treated as info, not real faults.
- On-demand or intentionally stopped containers are separated from possible container faults.
- If ZimaBrain can see an alert count but cannot safely parse the underlying detail, it should now say that clearly instead of acting like there is no alert.
- Added Files / AppData / media path diagnostics, including odd
/mediamount names, AppData symlink targets, read-only media, and ZimaOS media mirror path checks. - Added auditd ownership / permission checks.
- Improved answer trust labels so answers now show whether they are verified, partially verified, or guidance only.
- Added a small UI panel showing the new verifier improvements.
How to update:
cd /DATA/AppData/zimabrain-ce-flask-8601
git pull
docker compose build zimabrain-ce-flask-8601
docker compose up -d zimabrain-ce-flask-8601
Then check health:
curl -fsS http://127.0.0.1:8601/health && echo
docker ps --filter name=zimabrain-ce-flask-8601
Expected health output should include:
"ok": true
After updating, please test the same questions again:
What are the real alerts, container alerts, and info alerts?
What is the 1 Real Alert?
What are the 5 Info Only items?
This still depends on what the ZimaOS dashboard exposes in the local evidence, but the new build should be much clearer and should no longer hide the alert-count mismatch.
I also fixed the global question route. Questions like “what needs attention” now show VERIFIED FROM SAME-REPORT EVIDENCE when ZimaBrain finds real same-report verifier findings, instead of incorrectly falling back to guidance-only wording.