[Feature] Observability: structured logging, Prometheus metrics and OpenTelemetry tracing

[ElioNeto]

Overview

Add full observability to the vyx core and workers: structured logs, Prometheus metrics, and distributed tracing via OpenTelemetry.

Acceptance Criteria

• Core and workers emit structured JSON logs with: level, timestamp, service, correlation_id
• /metrics endpoint exposes Prometheus metrics:
  • Request count per route
  • Latency histogram per route
  • Error rate per route
  • Worker status (running/unhealthy/restarting)
• OpenTelemetry trace headers propagated from core to workers via message metadata
• Sample Grafana dashboard JSON provided in infra/grafana/
• Correlation ID generated per request and included in all downstream logs

References

• Spec §10 – Observability and Monitoring
• Roadmap – Phase 3

[ElioNeto]

Rastreamento de sub-tarefas

Esta issue é o épico de observabilidade. O progresso atual:

✅ Concluído

• [Feature] End-to-End Request Tracing (Correlation IDs) across Go Core and Workers #52 — End-to-End Request Tracing (Correlation IDs) — X-Request-Id gerado/propagado no Core, injetado no payload IPC e ecoado nas respostas HTTP.
• [Feature] Node.js Worker SDK: expose correlation ID via AsyncLocalStorage (#52) #64 — Node.js Worker SDK: AsyncLocalStorage com correlation_id — getCorrelationId() exportado do @vyx/worker.

🔧 Em andamento / próximas

• [Feature] Standardized Structured Logging across Polyglot Workers #53 — Standardized Structured Logging (JSON schema unificado para Go Core + Node.js + Python SDKs)
• [Feature] Python Worker SDK: expose correlation ID via contextvars (#52) #65 — Python Worker SDK: contextvars.ContextVar para correlation_id (desbloqueada — [Feature] End-to-End Request Tracing (Correlation IDs) across Go Core and Workers #52 está fechada)
• debt: logger do @vyx/worker não lê automaticamente do AsyncLocalStorage (#64 pendência) #67 — logger do @vyx/worker lendo req_id do AsyncLocalStorage (bloqueada por [Feature] Standardized Structured Logging across Polyglot Workers #53)

📋 Ainda sem issue dedicada

• Endpoint /metrics com Prometheus (request count, latency histogram, error rate por rota, status dos workers)
• OpenTelemetry trace headers propagados Core → workers
• Dashboard Grafana (infra/grafana/)

Sugestão: criar issues filhas para os critérios de Prometheus e OpenTelemetry antes de iniciar a implementação.