agentjido/jido_run/jido-otel

Jido OpenTelemetry

jido_otel is the upstream Hex package name.

Start Here

Use this skill when the task is about tracing, spans, correlation ids, or wiring Jido observability into an OpenTelemetry pipeline.

Good triggers:

• "Add OpenTelemetry tracing to this Jido app."
• "Implement the Jido tracer backend."
• "Review which spans and attributes this workflow should emit."
• "Turn the observability docs into a working tracer example."

Read references/builder-notes.md before coding when the task touches Jido.Observe, Jido.Observe.Tracer, or exporter configuration.

Primary Workflows

Implement tracing cleanly

• Start from the Jido observability hooks and span lifecycle instead of instrumenting random call sites.
• Keep tracer code focused on translating Jido metadata into OpenTelemetry spans and events.
• Preserve correlation ids and causation metadata when the docs expose them.

Configure exporters and runtime behavior

• Separate tracer implementation from exporter configuration.
• Keep production configuration explicit: endpoint, batching, redaction, and log level.
• Prefer safe defaults that do not break core Jido behavior when tracing is unavailable.

Turn docs into runnable examples

• Show one action or agent workflow producing spans.
• Include the Jido tracer configuration and one exporter setup.
• Keep examples small enough to inspect trace structure by hand.

Review boundaries

• Keep the observability facade in jido.
• Keep OpenTelemetry-specific integration in jido-otel.
• Keep product-specific dashboards and alerting in the application layer.

Build Checklist

• Confirm which Jido events should create spans.
• Propagate correlation metadata consistently.
• Decide what must be redacted before exporting.
• Add tests or examples for success, exception, and missing-exporter paths.

Boundaries

• Do not add direct OpenTelemetry dependencies to core jido behavior that the docs intentionally keep separate.
• Do not emit high-cardinality attributes without a clear need.
• Do not assume metrics or logs behavior beyond what the tracer integration documents.