@robota-sdk/agent-telemetry-otel 0.0.0-stage → 3.0.0-beta.94

package/CHANGELOG.md

@@ -0,0 +1,45 @@
+# @robota-sdk/agent-telemetry-otel
+
+## 3.0.0-beta.94
+### Minor Changes
+
+- ccf4194: `createOtelTracing({ tracer, mode, idGenerator })` records an agent run's provider calls and tool bodies
+  as OpenTelemetry spans on the host's tracer, under the host's span, named as `projectLivePromptTrace`
+  names them. Register its `plugin` once on the agent, and pass each run the `traceContext` and
+  `onExecutionEvent` from `startRun(parentSpanContext, { allowedOrigins, subprocessClasses })`.
+  `createStructuralIdGenerator()` is an `IdGenerator` for the host's `TracerProvider` so each span is
+  recorded under the span ID Robota sent in `traceparent`. `@opentelemetry/api` is a peer dependency.
+- fef3a50: New package `@robota-sdk/agent-telemetry-otel`. `projectLivePromptTrace(batch, { mode, agentName })`
+  maps one live prompt trace batch onto content-free OpenTelemetry span shapes without depending on an
+  OpenTelemetry SDK. `mode: 'agent'` keeps Robota's `agent.*` span and attribute names; `mode: 'gen_ai'`
+  uses the OpenTelemetry GenAI semantic conventions (`invoke_agent`, `chat`, `execute_tool`, `gen_ai.*`
+  attributes, `GEN_AI_SCHEMA_URL`) and keeps Robota-only facts such as timing and outcome as `agent.*`;
+  `mode: 'gen_ai/dup'` emits both attribute sets under GenAI span names for a gradual migration.
+  
+  `IToolBodyTraceEntry` gains an optional live-only `toolName`, which the GenAI projection uses for the
+  `execute_tool` span name and `gen_ai.tool.name`.
+- ccff111: Live telemetry can name its metrics with the OpenTelemetry GenAI semantic conventions. With
+  `PRODUCT_TELEMETRY_SEMCONV` set to `gen_ai`, the metrics export the `gen_ai.client.inference.*`
+  duration, token-usage and per-call token histograms, `gen_ai.execute_tool.duration` and
+  `gen_ai.invoke_agent.*` duration and call counts, and drop only `agent.provider.input_tokens` and
+  `agent.provider.output_tokens`, which the GenAI usage counters replace exactly. `gen_ai/dup` exports
+  every `agent.*` metric and the GenAI ones. The default, `agent`, exports exactly the metrics it did
+  before. Provider- and tool-derived GenAI metrics are exported only when no such child was omitted, and
+  usage that was not received is recorded as 0 under the `unknown` modality, as the conventions specify.
+  The GenAI metrics carry the requested model as those conventions require, and their inference points
+  also take the `agent.provider.id` and `agent.model.id` labels under the same
+  `PRODUCT_TELEMETRY_METRIC_ATTRIBUTES` opt-ins as the `agent.*` metrics (the model only in `gen_ai/dup`),
+  so a provider split of token totals survives the dropped counters. `PRODUCT_TELEMETRY_SEMCONV` now applies to
+  traces and metrics, and is refused only when neither is exported over otlp or console. Console metrics
+  print the same projection.
+  
+  `@robota-sdk/agent-telemetry-otel` adds `projectLivePromptMetricSamples`, which maps a live prompt trace
+  batch onto those GenAI metrics as plain, SDK-free samples with the recommended bucket boundaries, so
+  every host names and measures the same facts the same way. Its inference samples keep the label-safe
+  `agent.provider.id`, and `agent.model.id` in `gen_ai/dup`, as the provider span does.
+
+### Patch Changes
+
+- Updated dependencies [fef3a50]
+  - @robota-sdk/agent-interface-analytics@3.0.0-beta.94
+  - @robota-sdk/agent-core@3.0.0-beta.94