# AOC observability

Purpose
- Monitor Aggregation-Only ingestion for Concelier and Excititor.
- Provide deterministic metrics, traces, and logs for AOC guardrails.

Core metrics
- ingestion_write_total{source,tenant,result}
- ingestion_latency_seconds{source,tenant,phase}
- aoc_violation_total{source,tenant,code}
- ingestion_signature_verified_total{source,tenant,result}
- advisory_revision_count{source,tenant}
- verify_runs_total{tenant,initiator}
- verify_duration_seconds{tenant,initiator}

Alert guidance
- Violation spike: increase(aoc_violation_total[15m]) > 0 for critical sources.
- Stale ingestion: no growth in ingestion_write_total for > 60 minutes.
- Signature drop: rising ingestion_signature_verified_total{result="fail"}.

Health snapshot endpoint
- GET /obs/excititor/health returns ingest, link, signature, conflict status.
- Settings control warning and critical thresholds for lag, coverage, and conflict ratio.

Trace taxonomy
- ingest.fetch, ingest.transform, ingest.write
- aoc.guard for violations
- verify.run for verification jobs

Log fields
- traceId, tenant, source.vendor, upstream.upstreamId
- contentHash, violation.code, verification.window
- Correlation headers: X-Stella-TraceId, X-Stella-CorrelationId

Advisory AI chunk metrics
- advisory_ai_chunk_requests_total
- advisory_ai_chunk_latency_milliseconds
- advisory_ai_chunk_segments
- advisory_ai_chunk_sources
- advisory_ai_guardrail_blocks_total

Dashboards
- AOC ingestion health: sources overview, violations, signature rate, supersedes depth.
- Offline mode dashboard from offline snapshots.

Offline posture
- Metrics exporters write to local Prometheus snapshots in offline kits.
- CLI verification reports are hashed and archived.
- Dashboards support offline data sources.