Monitoring & Health
StellarTrail provides a suite of health endpoints and integrates with several observability tools to ensure reliable operation across all services.Health Endpoints
All health endpoints are served under/api/health. Public endpoints require no authentication. Admin endpoints require the x-admin-secret header.
Each endpoint returns a
status field (typically ok or degraded). The /ready endpoint returns 503 when the server cannot accept traffic; all other public endpoints return 200 unless a service is fully unavailable.
Background Queues
StellarTrail uses BullMQ backed by Redis for asynchronous job processing. Nine queues handle distinct workloads:
Queue health is classified as degraded when any single queue exceeds 10 failed or 50 waiting jobs, and unhealthy when a queue exceeds 100 failed or 500 waiting jobs.
Monitoring Tools
StellarTrail integrates the following tools for production observability:- Sentry — Error tracking and performance monitoring. Captures unhandled exceptions across both the server and client with full stack traces and request context.
- PostHog — Product analytics and feature flag management. Tracks user engagement, feature adoption, and conversion funnels.
- Pino — Structured JSON logging on the server. Provides request-level tracing with correlation IDs for debugging production issues.
- LangSmith — LLM observability and tracing. Records AI Coach prompt chains, token usage, and latency for each conversation turn.
- BullBoard — Admin dashboard for inspecting queue state. Provides a visual interface for monitoring job progress, retrying failed jobs, and draining queues.