feat!: Flatten JudgeResponse and EvalScore into new LDJudgeResult by jsonbailey · Pull Request #1284 · launchdarkly/js-core

jsonbailey · 2026-04-16T14:45:22Z

Summary

Replaces EvalScore and JudgeResponse with a flat LDJudgeResult interface (judgeConfigKey?, success, errorMessage?, sampled, score?, reasoning?, metricKey?)
Judge.evaluate() and evaluateMessages() now always return LDJudgeResult (never undefined). Sampling skip returns { sampled: false, success: false } instead of undefined.
sampled: true means the evaluation was sampled (actually run); sampled: false means it was skipped by sampling.
Replaces trackEvalScores and trackJudgeResponse with trackJudgeResult on both LDAIConfigTracker and LDGraphTracker. The new method guards on !result.sampled and !result.success — no event is emitted when either condition is true.
ChatResponse.evaluations type simplified from Promise<Array<JudgeResponse | undefined>> to Promise<LDJudgeResult[]>.
Public exports updated: EvalScore and JudgeResponse removed, LDJudgeResult added.

Mirrors the Python SDK change: launchdarkly/python-server-sdk-ai#132

Test plan

All 141 existing unit tests updated and passing
New tests for sampled=false (skipped) path in both config tracker and graph tracker
New test for success=false guard in both trackers
Lint clean

🤖 Generated with Claude Code

Note

Medium Risk
This is a breaking, cross-cutting API refactor that changes return types and tracking method names/semantics; downstream consumers may silently stop tracking judge metrics if they don’t propagate sampled/success or metricKey/score correctly.

Overview
Breaking API change: replaces JudgeResponse/EvalScore (evals map) with a flat LDJudgeResult (score, reasoning, metricKey, sampled, errorMessage, etc.), updating public exports accordingly.

Judge.evaluate()/evaluateMessages() now always return an LDJudgeResult (never undefined), returning explicit unsampled results when sampling skips execution and explicit error results for invalid config/provider failures; _parseEvaluationResponse now returns a single metric’s {score, reasoning} or undefined.

Tracking APIs are simplified: LDAIConfigTracker and LDGraphTracker replace trackEvalScores/trackJudgeResponse with trackJudgeResult, which only emits an event when sampled and success are true; TrackedChat and the direct-judge example/tests are updated to use the new shape and behavior.

^{Reviewed by Cursor Bugbot for commit 97b1ce8. Bugbot is set up for automated code reviews on this repo. Configure here.}

github-actions · 2026-04-16T14:46:56Z

@launchdarkly/js-sdk-common size report
This is the brotli compressed size of the ESM build.
Compressed size: 25623 bytes
Compressed size limit: 29000
Uncompressed size: 125843 bytes

github-actions · 2026-04-16T14:47:18Z

@launchdarkly/browser size report
This is the brotli compressed size of the ESM build.
Compressed size: 179375 bytes
Compressed size limit: 200000
Uncompressed size: 829982 bytes

github-actions · 2026-04-16T14:47:23Z

@launchdarkly/js-client-sdk-common size report
This is the brotli compressed size of the ESM build.
Compressed size: 37169 bytes
Compressed size limit: 38000
Uncompressed size: 204305 bytes

github-actions · 2026-04-16T14:47:25Z

@launchdarkly/js-client-sdk size report
This is the brotli compressed size of the ESM build.
Compressed size: 31655 bytes
Compressed size limit: 34000
Uncompressed size: 112792 bytes

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

…arseEvaluationResponse Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-authored-by: Jason Bailey <accounts@sidewaysgravity.com>

getTrackData() no longer accepts graphKey — it comes from the constructor. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

joker23

only a couple of optional nits

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

)

🤖 I have created a release *beep* *boop* --- <details><summary>browser: 0.1.16</summary> ## [0.1.16](browser-v0.1.15...browser-v0.1.16) (2026-04-21) ### Dependencies * The following workspace dependencies were updated * dependencies * @launchdarkly/js-client-sdk bumped from 4.6.0 to 4.6.1 </details> <details><summary>browser-telemetry: 1.0.32</summary> ## [1.0.32](browser-telemetry-v1.0.31...browser-telemetry-v1.0.32) (2026-04-21) ### Bug Fixes * correct typeof comparisons in browser SDK ([#1301](#1301)) ([f4bd636](f4bd636)) * **js-client-sdk:** better `undefined` handling ([#1303](#1303)) ([4818678](4818678)) ### Dependencies * The following workspace dependencies were updated * devDependencies * @launchdarkly/js-client-sdk bumped from 4.6.0 to 4.6.1 </details> <details><summary>js-client-sdk: 4.6.1</summary> ## [4.6.1](js-client-sdk-v4.6.0...js-client-sdk-v4.6.1) (2026-04-21) ### Bug Fixes * correct typeof comparisons in browser SDK ([#1301](#1301)) ([f4bd636](f4bd636)) * **js-client-sdk:** better `undefined` handling ([#1303](#1303)) ([4818678](4818678)) </details> <details><summary>react-sdk: 0.2.2</summary> ## [0.2.2](react-sdk-v0.2.1...react-sdk-v0.2.2) (2026-04-21) ### Dependencies * The following workspace dependencies were updated * dependencies * @launchdarkly/js-client-sdk bumped from ^4.6.0 to ^4.6.1 </details> <details><summary>server-sdk-ai: 0.17.0</summary> ## [0.17.0](server-sdk-ai-v0.16.8...server-sdk-ai-v0.17.0) (2026-04-21) ### ⚠ BREAKING CHANGES * Flatten JudgeResponse and EvalScore into new LDJudgeResult ([#1284](#1284)) * Add per-execution runId, at-most-once tracking, and cross-process tracker resumption ([#1270](#1270)) ### Features * Add per-execution runId, at-most-once tracking, and cross-process tracker resumption ([#1270](#1270)) ([fc25ab7](fc25ab7)) * Flatten JudgeResponse and EvalScore into new LDJudgeResult ([#1284](#1284)) ([aba1221](aba1221)) * Implement agent graph definitions ([#1282](#1282)) ([e7d08e5](e7d08e5)) * simplify evaluation schema to flat score/reasoning shape ([#1286](#1286)) ([c132e9f](c132e9f)) ### Bug Fixes * Add support for graph metric tracking ([#1269](#1269)) ([034a89d](034a89d)) </details> <details><summary>server-sdk-ai-langchain: 0.5.5</summary> ## [0.5.5](server-sdk-ai-langchain-v0.5.4...server-sdk-ai-langchain-v0.5.5) (2026-04-21) ### Dependencies * The following workspace dependencies were updated * devDependencies * @launchdarkly/server-sdk-ai bumped from ^0.16.8 to ^0.17.0 * peerDependencies * @launchdarkly/server-sdk-ai bumped from ^0.15.0 || ^0.16.0 to ^0.17.0 </details> <details><summary>server-sdk-ai-openai: 0.5.5</summary> ## [0.5.5](server-sdk-ai-openai-v0.5.4...server-sdk-ai-openai-v0.5.5) (2026-04-21) ### Dependencies * The following workspace dependencies were updated * devDependencies * @launchdarkly/server-sdk-ai bumped from ^0.16.8 to ^0.17.0 * peerDependencies * @launchdarkly/server-sdk-ai bumped from ^0.15.0 || ^0.16.0 to ^0.17.0 </details> <details><summary>server-sdk-ai-vercel: 0.5.5</summary> ## [0.5.5](server-sdk-ai-vercel-v0.5.4...server-sdk-ai-vercel-v0.5.5) (2026-04-21) ### Dependencies * The following workspace dependencies were updated * devDependencies * @launchdarkly/server-sdk-ai bumped from ^0.16.8 to ^0.17.0 * peerDependencies * @launchdarkly/server-sdk-ai bumped from ^0.15.0 || ^0.16.0 to ^0.17.0 </details> --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).  --- > [!NOTE] > **Medium Risk** > Primarily a version/changelog bump, but it publishes `@launchdarkly/server-sdk-ai` `0.17.0` with documented breaking API changes that can impact downstream consumers and provider peer dependency resolution. > > **Overview** > Bumps release versions across the monorepo via `.release-please-manifest.json`, updating `@launchdarkly/server-sdk-ai` to `0.17.0`, `@launchdarkly/js-client-sdk` to `4.6.1`, and related packages (`@launchdarkly/browser`, `@launchdarkly/react-sdk`, `@launchdarkly/browser-telemetry`, and AI provider packages) accordingly. > > Updates package metadata, changelogs, examples, and embedded SDK/wrapper version strings (e.g., `BrowserInfo` and `LDReactClient`) to reflect the new releases, including `server-sdk-ai`’s `0.17.0` breaking-change notes and provider peer dependency bumps to `^0.17.0`. > > <sup>Reviewed by [Cursor Bugbot](https://cursor.com/bugbot) for commit e7f8c09. Bugbot is set up for automated code reviews on this repo. Configure [here](https://www.cursor.com/dashboard/bugbot).</sup>  --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: jsonbailey <jbailey@launchdarkly.com> Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

jsonbailey commented Apr 16, 2026

View reviewed changes

Comment thread packages/sdk/server-ai/src/api/judge/Judge.ts Outdated

Comment thread packages/sdk/server-ai/src/api/judge/Judge.ts Outdated

Comment thread packages/sdk/server-ai/src/api/judge/types.ts Outdated

jsonbailey marked this pull request as ready for review April 16, 2026 15:10

jsonbailey requested a review from a team as a code owner April 16, 2026 15:10

jsonbailey and others added 6 commits April 16, 2026 10:15

feat!: Flatten JudgeResponse and EvalScore into new LDJudgeResult

fdfc425

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

refactor: use defaultJudgeResult factory and remove metricKey from _p…

b81c5c2

…arseEvaluationResponse Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chore: align LDJudgeResult property order to match spec

4ee6fbb

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

chore: combine sampled and success guards in trackJudgeResult

a639635

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Apply suggestions from code review

3c79377

Co-authored-by: Jason Bailey <accounts@sidewaysgravity.com>

fix: remove graphKey param from trackJudgeResult after upstream refactor

574ffd7

getTrackData() no longer accepts graphKey — it comes from the constructor. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

jsonbailey force-pushed the jb/aic-2200/simplify-judge-response branch from 6e54b30 to 574ffd7 Compare April 16, 2026 15:17

joker23 approved these changes Apr 16, 2026

View reviewed changes

Comment thread packages/sdk/server-ai/src/api/chat/TrackedChat.ts Outdated

Comment thread packages/sdk/server-ai/src/api/judge/types.ts Outdated

Comment thread packages/sdk/server-ai/src/api/judge/Judge.ts Outdated

jsonbailey and others added 2 commits April 16, 2026 11:50

refactor: remove defaultJudgeResult factory, use inline object literals

b33a85e

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

refactor: use object spread for successful return in Judge.evaluate

97b1ce8

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

joker23 approved these changes Apr 16, 2026

View reviewed changes

jsonbailey merged commit dd49a79 into feat/ai-sdk-next-release Apr 16, 2026
44 checks passed

jsonbailey deleted the jb/aic-2200/simplify-judge-response branch April 16, 2026 17:32

jsonbailey added a commit that referenced this pull request Apr 20, 2026

feat!: Flatten JudgeResponse and EvalScore into new LDJudgeResult (#1284

aba1221

)

github-actions Bot mentioned this pull request Apr 20, 2026

chore: release main #1298

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat!: Flatten JudgeResponse and EvalScore into new LDJudgeResult#1284

feat!: Flatten JudgeResponse and EvalScore into new LDJudgeResult#1284
jsonbailey merged 8 commits intofeat/ai-sdk-next-releasefrom
jb/aic-2200/simplify-judge-response

jsonbailey commented Apr 16, 2026 •

edited by cursor Bot

Loading

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

joker23 left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

jsonbailey commented Apr 16, 2026 • edited by cursor Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

joker23 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

jsonbailey commented Apr 16, 2026 •

edited by cursor Bot

Loading