effect-observability

Use Effect.log, Effect.logInfo, and related functions for log calls, then install a logger with Logger.layer, such as Logger.consoleJson for production. Attach queryable metadata with Effect.annotateLogs and control the minimum level via References.MinimumLogLevel.

Use Otlp.layerJson from effect/unstable/observability with a baseUrl and resource attributes, providing FetchHttpClient.layer. It exports traces, logs, and metrics to standard OTLP endpoints without an external OpenTelemetry SDK.

Yes. The PrometheusMetrics module in effect/unstable/observability formats metrics in Prometheus exposition format, and PrometheusMetrics.layerHttp registers a /metrics endpoint on your existing HttpRouter with optional path and prefix options.

Effect.withSpan wraps any effect in a named span, while Effect.fn with a string name automatically creates a span around the entire function body. Effect.fn is preferred for service methods since it combines definition and instrumentation.

Logger.layer replaces existing loggers by default, which can remove expected outputs. Pass { mergeWithExisting: true } to add loggers without replacing, and check that References.MinimumLogLevel is not filtering out your log level.

Use delta temporality for backends like Datadog or Dynatrace that expect per-interval values. Use the default cumulative temporality for Prometheus-style backends where values accumulate over the metric's lifetime.