You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Rename TypeSafe and LangChain Jev spans from evaluate to decide #25249
The TypeSafe integration (src/ai/typesafe/) makes the span for TypeSafeClient.systemOne(). The LangChain integration uses the same startEvaluateSpan() for TypeSafeClassifier runs and for Jev middleware calls inside createAgent / LangGraph. So one change covers all three paths.
Target span
Name: decide <model> (was evaluate <model>)
Op: gen_ai.decide (was gen_ai.evaluate)
gen_ai.operation.name: decide (was evaluate)
No other change. gen_ai.input.messages / gen_ai.output.messages keep type: 'evaluation'.
Scope
packages/server-utils/src/ai/typesafe/index.ts and src/ai/langchain/typesafe-classifier.ts. Rename startEvaluateSpan and related helpers to match.
Update the tests: node-integration-tests suites typesafe, langchain/v1, langgraph/v1; cloudflare-integration-tests suite langchain-typesafe-classifier; unit test test/ai/lib/tracing/langchain-invoke-agent.test.ts.
Activity
linear-code commentedon Oct 12, 2026
JS-3909
feat(server-utils): Record TypeSafe and LangChain Jev calls as gen_ai…
github-actions commentedon Oct 12, 2026
A PR closing this issue has just been released 🚀
This issue was referenced by PR #25270, which was included in the 11.7.0 release.