- Standardize invoke_agent pipeline span descriptions to use
invoke_agent (with optional functionId suffix) instead of Vercel SDK function names like generateText or generateObject. This aligns with how other AI integrations (e.g. LangGraph) name their agent spans.
- Unify all
.do* content generation span descriptions under a single generate_content prefix (e.g. generate_content mock-model-id) instead of using individual prefixes like generate_text, stream_text, generate_object, stream_object.
- Remove
addOriginToSpan helper and inline the setAttribute call directly.