App·Help

Troubleshooting

If something looks broken or empty, work through these in order.

1. The dashboard / a list page is empty

Most "empty" states are caused by one of three things, in order of likelihood:

  1. No source connected. Check Sources. If you see "No sources" or your sources are still pending, that's your answer.
  2. Source connected but first sync hasn't completed. Sync takes 30-60 seconds for small subscriptions. Wait, then refresh.
  3. Source synced but the underlying cloud has nothing to show. E.g. /anomalies needs a 28-day baseline; /drift needs an IaC Terraform source connected; /recommendations needs an investigation to have run. Each empty state explains its prerequisites — read the page.

2. Sign-in: AADSTS50011 redirect URI mismatch

The Entra app registration doesn't have your callback URI registered. Add https://<your-fqdn>/api/auth/callback/microsoft-entra-id (note microsoft-entra-id, not azure-ad) to the app registration's Authentication → Web → Redirect URIs.

3. Sign-in: generic "Server error"

Usually means NextAuth couldn't complete the callback because of a backend issue (DB unreachable, missing env var, schema mismatch). Stream the Container App logs:

az containerapp logs show -g rg-vintekh-pilot -n ca-vintekh-pilot-web --tail 200

Look for PrismaClientValidationError, P1001 (DB unreachable), or Configuration errors. Each has its own fix — most often a missing env var or a schema/code drift.

4. A specific page 500s

Same log command. The error stack will point at the failing query. Common shape:

Unknown field 'X' for select statement on model 'Y'
Unknown argument 'X'. Did you mean 'Z'?

These are schema/code drift bugs and one-line fixes — usually a rename or a missing field. File a bug with the error block; we can usually patch same-day.

5. Investigation: "blocked" or "needs review"

The AI orchestrator violated a contract guardrail or produced output the strict schema rejected. We never silently coerce — we surface. Re-run the investigation; if it persists, file a bug with the raw output (visible in an expandable details block on the investigation page).

6. Missing topology on a resource

Resource Graph relationship queries cover NSG / NIC / VNet peerings / PE / VMSS / LB backends today. Other join types are roadmap. Check Service coverage for the service — if topology is "roadmap" or "partial", that's expected for now.

7. Recommendations have low confidence

Confidence below 60 means the orchestrator couldn't gather enough evidence (sparse change events, missing topology, partial findings). The "Reasoning trail" panel on the investigation page will tell you what was missing — usually under Unknowns.

Still stuck

File a bug with: page URL, screenshot, the last 50 lines of relevant log, and what you expected to see vs. what you saw. We aim to triage within one business day during pilot.

Browse docs