evals(v4): OTEL/LangSmith + braintrust 3.x deps

[miguelg719]

Why

Retargets the LangSmith/OTEL eval-tracing work onto Stagehand 4.0. Stack 1/6 (deps).

What

• Add langsmith, @opentelemetry/{api,sdk-trace-base,sdk-trace-node,exporter-trace-otlp-proto}, @braintrust/otel; bump braintrust 0.4→3.x.
• @opentelemetry/api pinned catalog: (1.9.1) to share the SDK's global tracer.
• Fix braintrust-3.x wrapAISDK return-type in AISdkClientWrapped via Pick<typeof ai, …>.

Testing

typecheck + unit + build green.

Base: main

---

Summary by cubic

Adds OpenTelemetry and LangSmith tracing dependencies and upgrades braintrust to 3.x to prepare evals v4 for distributed tracing. No runtime behavior change; only dependency and type updates.

• Adds langsmith, @opentelemetry/{api,sdk-trace-base,sdk-trace-node,exporter-trace-otlp-proto}, @braintrust/otel; moves braintrust to runtime deps at ^3.21.0.
• Pins @opentelemetry/api via catalog: to share a single global tracer across the workspace.
• Disables native braintrust postinstall via workspace allowBuilds.braintrust: false.
• Fixes AISdkClientWrapped types for braintrust 3.x by deriving WrappedAI from ai and casting wrapAISDK(ai); public surface unchanged.
• Lockfile churn only. Migration: none — run pnpm install.

^{Written for commit 53a0364. Summary will update on new commits.}

[changeset-bot]

⚠️ No Changeset found

Latest commit: 53a0364

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[socket-security]

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	npm/​braintrust@​3.27.0
	npm/​langsmith@​0.5.26
	npm/​@​braintrust/​otel@​0.2.1
	npm/​@​opentelemetry/​sdk-trace-node@​2.10.0
	npm/​@​opentelemetry/​exporter-trace-otlp-proto@​0.220.0

View full report

[cubic-dev-ai]

All reported issues were addressed across 3 files

Architecture diagram

sequenceDiagram
    participant E as Eval Runner
    participant AI as AI SDK (ai module)
    participant BT as Braintrust SDK
    participant OTEL as OpenTelemetry SDK
    participant LS as LangSmith
    participant EXP as OTLP Exporter
    participant API as External Tracing API

    Note over E,API: Eval Tracing Infrastructure (v4)

    E->>AI: import { generateText, generateObject, streamText, streamObject }
    E->>AI: initialize AI SDK
    
    Note over E,BT: Wrapped AI SDK layer
    E->>BT: wrapAISDK(ai)
    BT->>BT: Create wrapped AISDK client
    BT-->>E: Wrapped client (typed via Pick)
    E->>AI: Call wrapped generateText/generateObject/streamText/streamObject

    Note over BT,OTEL: Braintrust OTEL integration
    BT->>OTEL: Get global tracer (catalog: 1.9.1)
    OTEL-->>BT: Shared tracer instance
    BT->>OTEL: Create spans for AI operations
    OTEL->>EXP: Export trace data via OTLP proto
    EXP->>API: Send spans to configured collector

    Note over BT,LS: LangSmith integration
    BT->>LS: Initialize LangSmith client
    LS->>LS: Configure trace processor
    BT->>LS: Send trace data
    LS->>API: Export to LangSmith backend

    alt Trace export success
        API-->>OTEL: Ack received
        API-->>LS: Ack received
    else Export failure
        API-->>OTEL: Error
        API-->>LS: Error
        Note over OTEL,LS: Failures non-blocking, eval continues
    end

    E->>E: Process eval results
    E-->>E: Return results (unchanged surface)

Loading

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>