feat: benchmark dashboard frontend and canonical data pipeline

[TheHiddenObserver]

Summary

Adds the statgpu benchmark dashboard, manifest-driven canonical data pipeline, audited source catalog and method-coverage matrix, current six-family CV evidence, deployed Vite/TypeScript/ECharts frontend, and permanent production cross-browser/accessibility QA.

Final status: MERGED_TO_MASTER_POST_MERGE_GREEN.

Merge result

PR head: c815b89974f529d5da13f4d8522c81746c4d20a3
previous master: f95447e8e1bf189f16ccc2d113b3acab2ae53dfe
master merge commit: 553f6b722a03b6d6870f8e2fe29b50bc11d38bfa
merged at: 2026-08-07T11:28:48Z

Completed staged work:

• benchmark: synchronize dashboard PR #76 with statgpu 0.2.4 master #90 / PR merge: synchronize benchmark dashboard with statgpu 0.2.4 #99 — synchronize with statgpu 0.2.4 master;
• benchmark: establish audited source catalog and canonical coverage matrix #100 / PR benchmark: establish source catalog and coverage foundation #110 — audited source catalog and coverage foundation;
• benchmark: add canonical cross-validation source and dashboard coverage #91 / PR benchmark: build canonical CV source pipeline and dashboard contract #111 — canonical six-family CV benchmark source and dashboard coverage;
• benchmark: complete dashboard production QA and accessibility gate #92 / PR benchmark: complete production dashboard QA and accessibility gate #113 — production-path, cross-browser, interaction, and accessibility QA.

Canonical bundle at merge

benchmark schema: 1.1.0
inventory version: 2.0
catalog version: 2.0
coverage matrix version: 1.0
discovered/classified JSON artifacts: 87
unclassified artifacts: 0
eligible sources: 9
registered sources: 9
available registered sources: 9
parsed registered sources: 9
eligible unregistered sources: 0
normalized runs: 1,796
minimum source date: 2026-06-01
generation ID: 51a83bca4a61c930b4001a6c4f47780b97ae75947366bc9cbd2ab9f5c9fbe915

The inventory describes the audited canonical bundle only and does not imply complete benchmark evidence for every maintained statgpu capability. Remaining family-specific gaps retain explicit ownership in the coverage matrix.

Canonical CV evidence

source_id: cv-benchmark-20260807-1347184c988d
canonical file: results/benchmark_frontend_sources/cv_benchmark_20260807.json
raw audit copy: results/cv_benchmark_candidate.json
sha256: 1347184c988d0f9648c8477d64752b646249282978cf28f65c165b391839bad2
measurement_git_sha: ad2cf88d1d443a53eeb5207c33c4ee4f25de2400
GPU: Tesla P100-SXM2-16GB

The source contains 22 framework/backend dispositions: 21 successful measured rows and one explicit LogisticRegressionCV Torch strict-CUDA failure. The failure remains canonical evidence without fabricated timing or CPU fallback; the implementation defect is tracked by #112.

Production QA

Permanent CI covers:

• committed nested deployment at /docs/assets/benchmarks/;
• Chromium, Firefox, and WebKit;
• the full CV filter cascade and deterministic upstream reset;
• visible keyboard focus, accessible names, skip navigation, scale/sort activation, and panel disclosure semantics;
• exact filter-synchronized timing/speedup tables as accessible chart alternatives;
• contrast, asset/JSON loading, metadata links, refresh, empty-state, and failed-CV-row behavior.

The production suite contains 5 product-level cases × 3 browser engines = 15 browser cases.

Pre-merge staging validation

All authoritative workflows on staging head c815b89974f529d5da13f4d8522c81746c4d20a3 passed:

Tests #1419 — success
Benchmark Frontend CI #490 — success
Maintenance compatibility #432 — success
Release notes validation #432 — success
Release package validation #438 — success

Fresh final review found no remaining blocking correctness, statistical-alignment, provenance, reproducibility, accessibility, deployment, documentation, maintainability, or CI issue. All inline review threads were resolved. Final review ID: 4882362644.

Post-merge master validation

All workflows triggered on exact master merge commit 553f6b722a03b6d6870f8e2fe29b50bc11d38bfa completed successfully:

Tests #1420 — success
Benchmark Frontend CI #491 — success
Maintenance compatibility #433 — success

master now points to the merge commit above. The dashboard integration is complete.

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 46c0722d2f

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.

[TheHiddenObserver]

Post-merge verification for PR #110 is complete on PR #76 head 69a3237ac471b55e5614527b9d522dd5ec77b847.

• Tests #1417 — success
• Benchmark Frontend CI #377 — success
• Maintenance compatibility #430 — success
• Release notes validation #430 — success
• Release package validation #436 — success
• audited inventory v2 remains deterministic with 85 classified artifacts, 0 unclassified artifacts, 8 eligible/registered/available/parsed sources, and no generated-asset drift
• no temporary workflow or unresolved review thread remains

The source catalog and coverage foundation tracked by #100 is complete. PR #76 should remain open for #91 and #92 before final merge to master.

[TheHiddenObserver]

Final integration review

Fresh full review completed on staging head c815b89974f529d5da13f4d8522c81746c4d20a3 against master baseline f95447e8e1bf189f16ccc2d113b3acab2ae53dfe.

Reviewed scope

Re-audited the final integrated source/catalog/coverage pipeline, canonical CV registration and non-success disposition contract, generated inventory/data provenance, frontend filter/chart/table identity rules, deployed nested-path assets, production cross-browser/accessibility layer, documentation guidance, and permanent CI configuration. I also rechecked the historical inline review set: all review threads are resolved, including prior timing/schema/provenance/chart-selection findings.

No temporary audit, integration, materialization, documentation, or marker workflow/script remains in the 151-file final diff.

Canonical evidence

• inventory v2: 87 classified JSON artifacts, 0 unclassified;
• 9 eligible / registered / available / parsed canonical sources;
• 1,796 normalized runs;
• current P100 CV source is SHA-protected and exposes 22 framework/backend dispositions;
• the sole LogisticRegressionCV Torch strict-CUDA failure remains explicit canonical evidence with no fabricated timing or CPU fallback and is owned by #112.

Product QA

The committed /docs/assets/benchmarks/ deployment has permanent production QA in Chromium, Firefox, and WebKit, plus the existing Chromium dev-server regression. Keyboard/focus naming, sortable controls, metric disclosures, explicit empty states, chart exact-value table fallbacks, text contrast, nested-path loading, refresh behavior, and the failed CV row are covered.

Current-head validation

All workflows triggered on the exact staging head are green:

• Tests #1419 — success;
• Benchmark Frontend CI #490 — success, including 6/6 jobs and production cross-browser QA;
• Maintenance compatibility #432 — success;
• Release notes validation #432 — success;
• Release package validation #438 — success.

GitHub reports the PR mergeable against master.

Conclusion

No blocking correctness, statistical-alignment, provenance, reproducibility, generated-asset, accessibility, deployment, documentation, maintainability, or CI finding remains in the intended PR #76 scope. #100, #91, and #92 are complete; the remaining LogisticRegressionCV Torch implementation defect is intentionally separated as #112 and does not invalidate the truthful benchmark evidence.

Final status: REVIEW_COMPLETE_MERGE_READY.

This review does not merge PR #76 into master; that should remain an explicit separate maintainer action.