Skip to content

Observability Index

Document ID: OBS-INDEX-001
File Path: docs/14-observability/index.md
Version: 1.0.0
Status: Active
Owner: Platform Operations Team
Last Updated: 2026-06-27


This document is the central navigation and architecture index for observability across the Wovyr AI Platform — logging, metrics, tracing, dashboards, and alerting. Observable by Default is a core principle.

This section is the platform-wide reference; subsystem docs (e.g. Tool Runtime Observability) implement it for their domain.


Logs ── what happened (structured, correlated)
Metrics ── how much / how fast (Prometheus)
Traces ── where time went (OpenTelemetry, end-to-end)
Events ── domain signals (Event Bus) for cost/usage/alerts

All four share correlation IDs (request_id, trace_id) so a symptom in one pillar links to the others.


ConcernTechnology
InstrumentationOpenTelemetry (traces, metrics, logs)
Metrics storePrometheus
DashboardsGrafana + native dashboard
Tracing backendOTLP-compatible (e.g. Tempo/Jaeger)
Log aggregationStructured log pipeline (e.g. Loki/ELK)

Per tech mapping and deployment.


DocumentResponsibility
logging.mdStructured logging standards
metrics.mdMetric taxonomy (RED/USE), cost metrics
tracing.mdDistributed tracing and correlation
dashboards.mdStandard dashboards and golden signals
alerting.mdAlert rules, SLOs, on-call

  1. Instrument everything — every service emits all three pillars.
  2. Correlate — one ID threads logs ↔ traces ↔ metrics.
  3. Golden signals — latency, traffic, errors, saturation everywhere.
  4. Cost is a signal — token/tool/memory spend is first-class.
  5. Observability ≠ audit — see audit for security-grade records.

Observability is for operations (may be sampled, shorter retention). Security audit (13-security/audit.md) is integrity-protected and compliance-retained. They are separate pipelines.



VersionDateDescription
1.0.02026-06-27Initial Observability Index