[OPIK-6838] [FE] fix: link Playground prompt versions to experiments and traces by awkoy · Pull Request #7017 · comet-ml/opik

awkoy · 2026-06-10T13:34:47Z

Details

Experiments and traces created from the Playground over a regular Dataset (or via a direct run) silently lost their Prompt Library version link, while runs over a Test Suite kept it. Root cause: the dataset/"log" path collected only per-message TEXT prompt versions and ignored the loaded CHAT prompt version (loadedChatPromptVersionId), whereas the test-suite/"execute" path collected both. This produced an empty "Prompt commit" column, dataset experiments missing from a prompt's Experiments tab, and no prompt reference in the single-experiment view.

This PR removes that drift by introducing one shared version collector used by both paths, surfaces the prompt-version link in the Configuration tab, and keeps the trace↔prompt reference even after a loaded prompt is edited. Frontend-only — no backend, migrations, or v1 changes.

Extract collectPromptVersionRefs + buildPromptLibraryMetadata into a shared, unit-tested src/api/playground/promptLinkage.ts (single source of truth); reuse it in both the execute and log paths.
Dataset/log path now links CHAT prompt versions → fixes the empty "Prompt commit" column and dataset experiments missing from the prompt's Experiments tab.
Add a prompt-version link in the single-experiment Configuration tab (deep-links to the experiment's exact version via activeVersionId).
Trace linkage (metadata.opik_prompts, a frontend-only field) now keeps the reference for edited prompts with a modified flag, plus a "modified" badge on the trace's Prompts card; TEXT path anchors to the message's own version instead of latest.

Change checklist

User facing
Documentation update

Issues

OPIK-6838

AI-WATERMARK

AI-WATERMARK: yes

Tools: Claude-Code
Model: Opus-4.8
Scope: All files
Human verification: Yes — reproduced all 4 issues as broken on main, then verified each fixed in the running app (local, OpenAI provider, Chrome).

Testing

Commands (from apps/opik-frontend):

npm run typecheck — clean (project-wide)
npm run lint — 0 warnings (project-wide)
npm run test -- src/api/playground — pass, including new promptLinkage.test.ts (9 cases)
npx vitest run (full suite) — 1751 passed. The only failures (23 in src/lib/ls-migrations/*) are a pre-existing happy-dom localStorage.clear env issue, unrelated to this change (they fail in isolation and touch none of the changed files).

Manual UI (local process mode, OpenAI provider, Chrome) — loaded a CHAT library prompt into the Playground and ran an experiment over a regular Dataset:

[CM-7070] llm sdk allow users to logs basic llm prompts and answers #1 "Prompt commit" column populated with the version commit (was -).
[NA] debug deploy #3 the dataset experiment now appears under the prompt's Experiments tab.
Update README.md #2 the single-experiment Configuration tab shows a prompt-version link.
Update log_prompt docstring #4 the resulting trace carries opik_prompts (with the new modified field) and renders the prompt card on the Prompts tab.
Confirmed all four reproduce as broken on main.

Not run: a live screenshot of the modified: true badge — local ClickHouse trace-list ingestion lag prevented surfacing the edited-run trace in time. The modified flag is confirmed present in trace metadata and the badge rendering is covered by code review.

Documentation

N/A — no user-facing docs change.

🤖 Generated with Claude Code

…e path (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

… (OPIK-6838 #1,#3)

…ed flag (OPIK-6838 #4) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…PIK-6838 #4)

…838 #2) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…on in Configuration tab (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…er (OPIK-6838) Addresses the Baz review comment plus the multi-agent final-review findings: - Dedupe the version-resolution logic shared by the CHAT and TEXT paths in useHydratePromptMetadata into a single tested helper. - Resolve the embedded latest_version without an extra fetch when no explicit version is requested (restores pre-refactor perf parity for the TEXT path). - Pass the resolved PromptVersion straight to buildPromptLibraryMetadata instead of re-spreading its fields at both call sites. - Add unit tests for resolvePromptVersionForLink, metadata pass-through, and multi-message collection.

baz-reviewer · 2026-06-10T14:30:35Z

+export const resolvePromptVersionForLink = async (
+  promptData: { latest_version?: PromptVersion } | undefined,
+  explicitVersionId: string | undefined,
+  fetchPromptVersion: FetchPromptVersion,
+): Promise<PromptVersion | undefined> => {


resolvePromptVersionForLink duplicates the version-resolution/fetch logic in v1 useHydratePromptMetadata, so should we reuse this helper there too or extract a shared function?

Want Baz to fix this for you? Activate Fixer

Commit 2c53828 addressed this comment by extracting the prompt-version linkage logic into shared helpers in promptLinkage.ts, which matches the suggested “extract a shared function” refactor. That removes the duplicated version-selection logic from the old per-path implementation.

github-actions · 2026-06-10T14:43:03Z

🔄 Test environment deployment process has started

Phase 1: Deploying base version 2.0.61-5585 (from main branch) if environment doesn't exist
Phase 2: Building new images from PR branch fix/opik-6838-prompt-linkage
Phase 3: Will deploy newly built version after build completes

You can monitor the progress here.

…vation (OPIK-6838) Replace the compound ternary (compare-guard + optional chaining + .length truthiness + empty fallback) with a clear compare-vs-single branch; the consumer maps the array, so an empty array renders nothing.

baz-reviewer · 2026-06-10T14:48:01Z

  const isCompare = experimentsIds.length > 1;

+  // Prompt-version links only make sense for a single experiment; in compare
+  // mode we don't show them. An empty array simply renders nothing.
+  const promptVersions = isCompare ? [] : experiments[0]?.prompt_versions ?? [];
+
  const [columnsWidth, setColumnsWidth] = useLocalStorageState<


promptVersions/NavigationTag prompt-link rendering duplicates CompareExperimentsDetails.tsx, should we extract a shared helper/component so both tabs stay in sync?

Want Baz to fix this for you? Activate Fixer

CometActions · 2026-06-10T14:52:16Z

✅ Test environment is now available!

To configure additional Environment variables for your environment, run [Deploy Opik AdHoc Environment workflow] (https://github.com/comet-ml/comet-deployment/actions/workflows/deploy_opik_adhoc_env.yaml)

Access Information

URL: https://pr-7017.dev.comet.com
Cluster: comet-ml-development
Namespace: pr-7017
Version: 2.0.61-7017-merge-2259
Application logs: View in Grafana

The deployment has completed successfully and the version has been verified.

…d const (OPIK-6838)

baz-reviewer · 2026-06-10T14:54:23Z

+    const versionRefs = collectPromptVersionRefs(prompt);
+    const promptVersions = versionRefs.length
+      ? versionRefs.map((ref) => ({ id: ref.id, prompt_id: ref.promptId }))
+      : undefined;


Lines 39-42 duplicate the collectPromptVersionRefs(prompt) -> prompt_versions/promptLibraryVersions mapping from usePromptDatasetItemCombination.ts; should we move this into promptLinkage.ts as something like buildPromptVersionsPayload so both paths share the same {id, prompt_id?} shape?

Want Baz to fix this for you? Activate Fixer

aadereiko

LGTM :)

Follow-up to PR #7017. Review feedback on the prompt-linkage work flagged that experiment surfaces show the raw prompt commit hash, which users don't recognize. Show the human-readable version (version_number, e.g. v3) next to the prompt name, mirroring the already-approved Prompt Library experiments tab, and fall back to the commit hash only when no version number is available (null for masks). Surfaces updated (all v2, display-only — no data-layer/BE change): - ExperimentsPage/GeneralDatasetsTab: "Prompt commit" column -> "Prompt", rendering "<name> (<version>)" via version_label. - CompareExperimentsPage/ConfigurationTab: NavigationTag label built from version_number ?? commit instead of commit. - dashboards/ExperimentsLeaderboardWidget: same column fix (caught in review; identical commit-hash issue on the dashboard leaderboard). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…sh in experiment views (#7031) * [OPIK-6838] [FE] fix: show prompt name + version instead of commit hash Follow-up to PR #7017. Review feedback on the prompt-linkage work flagged that experiment surfaces show the raw prompt commit hash, which users don't recognize. Show the human-readable version (version_number, e.g. v3) next to the prompt name, mirroring the already-approved Prompt Library experiments tab, and fall back to the commit hash only when no version number is available (null for masks). Surfaces updated (all v2, display-only — no data-layer/BE change): - ExperimentsPage/GeneralDatasetsTab: "Prompt commit" column -> "Prompt", rendering "<name> (<version>)" via version_label. - CompareExperimentsPage/ConfigurationTab: NavigationTag label built from version_number ?? commit instead of commit. - dashboards/ExperimentsLeaderboardWidget: same column fix (caught in review; identical commit-hash issue on the dashboard leaderboard). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * [OPIK-6838] [FE] refactor: extract shared formatPromptVersionLabel helper Address review feedback (baz): the "<name> (<version>)" formatting was duplicated across the three surfaces. Centralize it in a single tested helper in lib/experiments.ts and reuse it from the Experiments table column, the single-experiment Configuration tab, and the dashboard leaderboard widget so the display rule changes in one place. Also adds unit tests for the version-number/commit fallback and the no-version case. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

…and traces (#7017) * feat(playground): add shared prompt-version linkage helpers (OPIK-6838) * refactor(playground): reuse shared prompt-version collector in execute path (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(playground): link CHAT prompt versions on dataset experiment runs (OPIK-6838 #1,#3) * fix(playground): keep trace prompt link on edited prompts with modified flag (OPIK-6838 #4) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * feat(traces): show modified badge on edited playground prompt link (OPIK-6838 #4) * feat(experiments): link prompt version from Configuration tab (OPIK-6838 #2) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(playground): anchor TEXT prompt version + link experiment's version in Configuration tab (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * refactor(playground): extract shared resolvePromptVersionForLink helper (OPIK-6838) Addresses the Baz review comment plus the multi-agent final-review findings: - Dedupe the version-resolution logic shared by the CHAT and TEXT paths in useHydratePromptMetadata into a single tested helper. - Resolve the embedded latest_version without an extra fetch when no explicit version is requested (restores pre-refactor perf parity for the TEXT path). - Pass the resolved PromptVersion straight to buildPromptLibraryMetadata instead of re-spreading its fields at both call sites. - Add unit tests for resolvePromptVersionForLink, metadata pass-through, and multi-message collection. * refactor(experiments): simplify Configuration tab promptVersions derivation (OPIK-6838) Replace the compound ternary (compare-guard + optional chaining + .length truthiness + empty fallback) with a clear compare-vs-single branch; the consumer maps the array, so an empty array renders nothing. * refactor(playground): hoist prompt_versions request mapping to a named const (OPIK-6838) --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>

…sh in experiment views (#7031) * [OPIK-6838] [FE] fix: show prompt name + version instead of commit hash Follow-up to PR #7017. Review feedback on the prompt-linkage work flagged that experiment surfaces show the raw prompt commit hash, which users don't recognize. Show the human-readable version (version_number, e.g. v3) next to the prompt name, mirroring the already-approved Prompt Library experiments tab, and fall back to the commit hash only when no version number is available (null for masks). Surfaces updated (all v2, display-only — no data-layer/BE change): - ExperimentsPage/GeneralDatasetsTab: "Prompt commit" column -> "Prompt", rendering "<name> (<version>)" via version_label. - CompareExperimentsPage/ConfigurationTab: NavigationTag label built from version_number ?? commit instead of commit. - dashboards/ExperimentsLeaderboardWidget: same column fix (caught in review; identical commit-hash issue on the dashboard leaderboard). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * [OPIK-6838] [FE] refactor: extract shared formatPromptVersionLabel helper Address review feedback (baz): the "<name> (<version>)" formatting was duplicated across the three surfaces. Centralize it in a single tested helper in lib/experiments.ts and reuse it from the Experiments table column, the single-experiment Configuration tab, and the dashboard leaderboard widget so the display rule changes in one place. Also adds unit tests for the version-number/commit fallback and the no-version case. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

awkoy and others added 7 commits June 10, 2026 12:21

feat(playground): add shared prompt-version linkage helpers (OPIK-6838)

eda8d27

refactor(playground): reuse shared prompt-version collector in execut…

a1c5a39

…e path (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(playground): link CHAT prompt versions on dataset experiment runs…

200fd0c

… (OPIK-6838 #1,#3)

fix(playground): keep trace prompt link on edited prompts with modifi…

5994cfc

…ed flag (OPIK-6838 #4) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

feat(traces): show modified badge on edited playground prompt link (O…

0a4e212

…PIK-6838 #4)

feat(experiments): link prompt version from Configuration tab (OPIK-6…

3907f6b

…838 #2) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

fix(playground): anchor TEXT prompt version + link experiment's versi…

498c630

…on in Configuration tab (OPIK-6838) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

awkoy requested a review from a team as a code owner June 10, 2026 13:34

github-actions Bot assigned awkoy Jun 10, 2026

github-actions Bot added Frontend typescript *.ts *.tsx labels Jun 10, 2026

baz-reviewer Bot reviewed Jun 10, 2026

View reviewed changes

Comment thread apps/opik-frontend/src/v2/pages/PlaygroundPage/useHydratePromptMetadata.ts Outdated

baz-reviewer Bot reviewed Jun 10, 2026

View reviewed changes

awkoy added the test-environment Deploy Opik adhoc environment label Jun 10, 2026

baz-reviewer Bot reviewed Jun 10, 2026

View reviewed changes

refactor(playground): hoist prompt_versions request mapping to a name…

2c53828

…d const (OPIK-6838)

baz-reviewer Bot reviewed Jun 10, 2026

View reviewed changes

aadereiko approved these changes Jun 10, 2026

View reviewed changes

awkoy merged commit b332629 into main Jun 10, 2026
10 checks passed

awkoy deleted the fix/opik-6838-prompt-linkage branch June 10, 2026 15:39

awkoy mentioned this pull request Jun 11, 2026

[OPIK-6838] [FE] fix: show prompt name + version instead of commit hash in experiment views #7031

Merged

2 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[OPIK-6838] [FE] fix: link Playground prompt versions to experiments and traces#7017

[OPIK-6838] [FE] fix: link Playground prompt versions to experiments and traces#7017
awkoy merged 10 commits into
mainfrom
fix/opik-6838-prompt-linkage

awkoy commented Jun 10, 2026

Uh oh!

Uh oh!

baz-reviewer Bot Jun 10, 2026

Uh oh!

baz-reviewer Bot Jun 10, 2026

Uh oh!

github-actions Bot commented Jun 10, 2026

Uh oh!

baz-reviewer Bot Jun 10, 2026

Uh oh!

CometActions commented Jun 10, 2026

Uh oh!

baz-reviewer Bot Jun 10, 2026

Uh oh!

aadereiko left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

awkoy commented Jun 10, 2026

Details

Change checklist

Issues

AI-WATERMARK

Testing

Documentation

Uh oh!

Uh oh!

baz-reviewer Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

baz-reviewer Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot commented Jun 10, 2026

Uh oh!

baz-reviewer Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

CometActions commented Jun 10, 2026

Access Information

Uh oh!

baz-reviewer Bot Jun 10, 2026

Choose a reason for hiding this comment

Uh oh!

aadereiko left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants