Runbooks & troubleshooting (operations)
AdminThis page bundles common operational problems and points to the matching detail runbooks. The approach is always: symptom → first check → detail runbook.
Sign-in fails (no OTP code)
Section titled “Sign-in fails (no OTP code)”First check: SMTP status in the System settings (last health check). Often a wrong host/port/sender, or an unreachable SMTP service in server operation. → Set up OTP sign-in
Search returns no/poor hits
Section titled “Search returns no/poor hits”First check: the active vector backend. Without Qdrant/sqlite-vec, BM25 runs (lower hit quality). Wait for the reindex after a restore/import. → Set up Qdrant · RAG & search
PII is not pseudonymized
Section titled “PII is not pseudonymized”First check: Presidio health. On failure no pseudonymization happens (no local fallback). → Set up Presidio
Model/MCP does not respond
Section titled “Model/MCP does not respond”First check: API keys and provider status. After a restore/peer sync, keys must be correctly re-encrypted. → LLM API keys · MCP servers
Health probe red / container restarts
Section titled “Health probe red / container restarts”First check: /ready and the logs. A “ready” check (e.g. database) is failing. → Monitoring & health checks
Columns missing after update / startup error
Section titled “Columns missing after update / startup error”The cause is usually the two-spot rule (EF migration without a matching DbInitializer entry). On startup only DbInitializer.InitializeAsync runs. → Updates & version changes
Restore fails
Section titled “Restore fails”First check: passphrase correct? App version ≥ backup version? → Backup & restore
Clients have a diverging configuration
Section titled “Clients have a diverging configuration”First check: the diff in the Connections tab; is the host reachable? → Peer sync / server mode in operation
Diagnostic basics
Section titled “Diagnostic basics”- Audit logs (knowledge write operations, PII). → Audit log (PII)
- Usage (tokens/cost/latency) for anomalies. → Usage statistics
- Telemetry (OTLP) and health endpoints for the infrastructure. → Monitoring & health checks