J.A.R.V.I.S.

EVIDENCE-BACKED ROOT CAUSE · BLIND-TESTED

0raw bad events

›

0deduped alerts

›

0open incidents

connecting… Logins/min · success Payments/min – · success – user-db connections –

Service topology

healthy anomalous (symptom) likely root cause dashed = inferred from callers

Incidents

✓

All systems nominal. Watching live telemetry…

1Collect
logs, latencies, health checks and Postgres stats from every service

2Correlate
which broken service explains the others; each unexplained one is an incident

3Score
3A + 2O + 2L + 1C − 3X picks the root cause, with evidence for every point

4Prove
break something in the blind test lab and reveal the hidden truth

Log-file upload

.jsonl lines are read as-is. Plain text uses
<date> <time> <LEVEL> [<service>] <message>

Lines that do not match are kept with service=unknown, never dropped. Try samples/db_pool_incident.log.

External alerts (webhook)

curl -X POST localhost:8000/webhook/alert \
  -H "Content-Type: application/json" \
  -d '{"alerts":[{"status":"firing","labels":{"service":"payment-service",
      "alertname":"HighLatency","severity":"critical"},
      "startsAt":"2026-11-09T14:03:00Z"}]}'