benchmark: synchronize dashboard PR #76 with statgpu 0.2.4 master

[TheHiddenObserver]

Context

PR #76 contains the benchmark dashboard, manifest-driven data pipeline, frontend, generated deployment assets, and CI contracts. Its current head is based on an older master snapshot and GitHub does not currently consider it mergeable.

The released baseline is statgpu 0.2.4 at commit:

0aeeb95b60e3e274053b8f1b6427ae50c8eec015

This issue tracks synchronization only. It is the first executable item in the roadmap introduced by PR #89.

Goal

Bring feature/benchmark-frontend-dashboard onto the current 0.2.4 master baseline without changing the dashboard schema or adding new benchmark families.

Scope

• merge or rebase current master into the PR feat: benchmark dashboard frontend and canonical data pipeline #76 branch;
• resolve conflicts in tests, workflows, documentation, package layout, benchmark sources, parsers, generated JSON, and deployed assets;
• preserve the existing manifest SHA256, canonical run identity, semantic validation, transactional generation, and no-fabrication contracts;
• regenerate:
  • benchmark_data.json;
  • parse_report.json;
  • source_inventory.json;
  • docs/assets/benchmarks/;
• update PR feat: benchmark dashboard frontend and canonical data pipeline #76 metadata and status after synchronization.

Non-goals

• no new benchmark source or method family;
• no schema version change unless required solely to preserve an existing contract after the merge, in which case it must be split into a follow-up issue;
• no URL-persisted state, mobile redesign, virtualization, or bundle partitioning;
• no merge of PR feat: benchmark dashboard frontend and canonical data pipeline #76 into master as part of this issue.

Required validation

• project tests on Python 3.9, 3.10, 3.11, and 3.12;
• benchmark parser/data tests on maintained Python versions;
• strict deterministic generator check;
• JSON Schema and semantic validation;
• TypeScript typecheck;
• Vite production build;
• deterministic staleness check for generated data and deployed assets;
• Playwright Chromium E2E;
• clean final tree with no temporary workflows, patch scripts, diagnostics, or marker files.

Acceptance criteria

• PR feat: benchmark dashboard frontend and canonical data pipeline #76 head contains current 0.2.4 master history.
• GitHub reports the PR mergeable or any remaining blocker is explicitly documented and scoped.
• No canonical source hash or run identity changes without an explained source or contract change.
• The generated three-file bundle and deployed assets are current and deterministic.
• All required Python, frontend, build, staleness, and browser gates pass.
• No new benchmark family is introduced in the synchronization diff.
• PR feat: benchmark dashboard frontend and canonical data pipeline #76 body records the synchronized base commit and final validation runs.

Dependencies

• Roadmap reconciliation PR docs: rebuild development roadmap for 0.2.4 #89 should merge first or remain the documented planning reference during implementation.

Follow-up

After synchronization, current CV data is tracked separately and dashboard product QA is tracked separately.

[TheHiddenObserver]

Final implementation plan — reviewed and fixed

Impact classification

Active axes:

• benchmark/data provenance;
• CI and generated-artifact contracts;
• documentation/navigation;
• branch/history integration.

Inactive axes:

• no public estimator API change;
• no backend/device behavior change;
• no loss, penalty, solver, CV, inference, or formula implementation change;
• no new performance claim.

Scope and branch strategy

• Synchronize PR feat: benchmark dashboard frontend and canonical data pipeline #76 with current master commit f95447e8e1bf189f16ccc2d113b3acab2ae53dfe.
• Create agent/sync-benchmark-dashboard-v0.2.4 from PR feat: benchmark dashboard frontend and canonical data pipeline #76 head 10874915f05dcae8f0124870e7d10de31848d42b.
• Produce a real two-parent merge commit whose first parent is the PR feat: benchmark dashboard frontend and canonical data pipeline #76 head and whose second parent is current master.
• Open a draft stacked PR with base feature/benchmark-frontend-dashboard.
• Do not merge PR feat: benchmark dashboard frontend and canonical data pipeline #76 into master in this phase.

Pre-merge invariants

Record the PR #76 baseline before resolving conflicts:

• schema 1.1.0;
• 8 registered sources;
• 8 parsed sources;
• 0 skipped sources;
• 1,774 normalized runs;
• 36 models;
• minimum source date 2026-06-01;
• manifest source paths and SHA256 values;
• generation IDs and hashes of the three committed data files;
• committed deployment-asset inventory.

These values may change only when deterministic rebuilding under the same registered inputs and contracts proves that a prior committed artifact was stale or invalid. Any such drift must be explained row-by-row/source-by-source rather than accepted as incidental merge output.

Confirmed overlap inventory

The old PR #76 base is a4879fb4d9fb183efc01f147cd2cc501691f28c4. Comparing it with current master identifies five paths changed on both sides:

1. .gitignore
2. docs/en/README.md
3. docs/en/guides/benchmarks.md
4. docs/cn/guides/benchmarks.md
5. pyproject.toml

Conflict policy:

• .gitignore: union the current master exclusions with dashboard/frontend exclusions; remove duplicates and do not weaken artifact protections.
• docs/en/README.md: retain current master navigation and capability wording, then add the dashboard guide link in the appropriate benchmark/navigation section.
• EN/CN benchmark guides: preserve all current 0.2.4 benchmark guidance and integrate the dashboard entry without restoring stale performance claims.
• pyproject.toml: preserve current 0.2.4 package metadata, dependencies, pytest settings, and release configuration; add only the dashboard-specific test/config entries still required by PR feat: benchmark dashboard frontend and canonical data pipeline #76.

All non-overlapping PR #76 files are retained as PR #76 content; all non-overlapping master files are inherited from current master.

Inherited but unregistered benchmark files

Current master adds results/benchmark_frontend_sources/coxph_boundary_workspace_pr80_20260728.json after the old PR #76 base.

For this synchronization phase:

• preserve the file because it is part of master history;
• do not add it to frontend_sources.json;
• do not add a parser, source registration, model family, or dashboard row for it;
• verify the generator and tests tolerate registered-source-only canonical mode;
• require the canonical inventory to remain at the existing eight registered sources unless a pre-existing artifact defect is independently proven.

Registering this or any other new source belongs to a separate issue/PR.

Merge construction and resolution

1. Build the merged tree from current master, overlaying the PR feat: benchmark dashboard frontend and canonical data pipeline #76 file surface.
2. Manually resolve the five confirmed overlapping paths according to the policies above.
3. Preserve current master behavior for estimator, solver, package, test, release, and maintenance changes.
4. Preserve PR feat: benchmark dashboard frontend and canonical data pipeline #76 behavior for schema v1.1.0, manifest SHA256, canonical identities, semantic validation, transactional generation, no-fabrication rules, frontend behavior, and deployment assets.
5. Audit the resulting tree for accidental loss of master files and accidental introduction of temporary files.

Generated-data and asset validation

• Verify every registered source exists and matches its manifest SHA256.
• Run strict deterministic generation for:
  • frontend/public/data/benchmark_data.json;
  • frontend/public/data/parse_report.json;
  • frontend/public/data/source_inventory.json.
• Validate schema, semantic constraints, canonical run IDs, source SHA propagation, chart-cell speedup-reference identity, finite metrics, and date-time formats.
• Rebuild docs/assets/benchmarks/ from the synchronized frontend.
• Compare generated inventory with the recorded pre-merge baseline.
• Reject unexplained source/hash/run/model/generation drift and never fabricate or infer missing rows.

Validation gates

Hosted/project gates:

• project Tests workflow on Python 3.9, 3.10, 3.11, and 3.12;
• Maintenance compatibility workflow;
• Benchmark Frontend CI Python data jobs on 3.9 and 3.11;
• strict generator check;
• JSON Schema plus semantic validation;
• TypeScript typecheck;
• Vite production build;
• deterministic data/deployment staleness job;
• Playwright Chromium E2E.

Tree and scope gates:

• merge commit contains both required parents;
• stacked diff introduces no new source registration, parser family, schema version, model family, or frontend feature;
• no temporary workflow, patch script, diagnostic, marker, Node dependency, test-results, or Playwright report is committed;
• PR feat: benchmark dashboard frontend and canonical data pipeline #76 remains unmerged.

Review-fix protocol

Plan review

The plan is complete only after checking it against:

• issue benchmark: synchronize dashboard PR #76 with statgpu 0.2.4 master #90 acceptance criteria;
• dev/AGENTS.md;
• .claude/workflows/new-module-dev.md hard exit/review rules;
• PR feat: benchmark dashboard frontend and canonical data pipeline #76 schema/parser/aggregation/rollout contracts;
• current master-to-old-base overlap inventory.

Plan review result after this revision: no remaining blocking plan finding.

Implementation review

After creating the stacked PR:

1. Review the merge topology, full stacked diff, five manual resolutions, canonical artifacts, CI, and PR metadata.
2. Fix every CRITICAL/HIGH and every in-scope correctness or reproducibility finding.
3. Rerun affected gates.
4. Perform a fresh review against the new head.
5. Repeat until the latest fresh review reports no new in-scope issue.

Completion and issue state

• Keep the stacked PR draft while implementation review is active.
• When clean, mark it ready for review and update PR feat: benchmark dashboard frontend and canonical data pipeline #76 with:
  • synchronized master commit;
  • stacked/merge commit;
  • canonical bundle inventory;
  • validation run IDs;
  • explicit statement that no new benchmark family was introduced.
• Merge of the stacked PR into feature/benchmark-frontend-dashboard still requires a separate explicit merge action.
• Close benchmark: synchronize dashboard PR #76 with statgpu 0.2.4 master #90 only after that stacked PR is merged into the PR feat: benchmark dashboard frontend and canonical data pipeline #76 branch and PR feat: benchmark dashboard frontend and canonical data pipeline #76 becomes mergeable, or an external remaining blocker is precisely documented.

Hard exit target

Target status: COMPLETE for the synchronization PR itself when all local/hosted active gates and review are clean. PR #76 product integration into master remains intentionally out of scope.

[TheHiddenObserver]

Synchronization implementation complete — stacked PR #99 ready

Issue #90 has completed its plan, implementation, and review-fix phases up to the required merge boundary.

Final state

• stacked PR: merge: synchronize benchmark dashboard with statgpu 0.2.4 #99;
• target: feature/benchmark-frontend-dashboard (PR feat: benchmark dashboard frontend and canonical data pipeline #76 staging branch);
• final head: c394accf713d798ef8adf6474a75ebeb987e8ce1;
• synchronized master: f95447e8e1bf189f16ccc2d113b3acab2ae53dfe;
• two-parent synchronization merge: 8c72e94e3c8b1a1527574ce1fe3b220acb09c562;
• PR merge: synchronize benchmark dashboard with statgpu 0.2.4 #99 is open, mergeable, and Ready for review.

Canonical bundle

• schema 1.1.0;
• minimum source date 2026-06-01;
• catalog 40, with 8 eligible/registered/available/parsed sources;
• 0 skipped sources;
• 1,774 normalized runs;
• 36 models;
• generation ID 7437ea045e8161b4090598bdfb8062068514a5914f5d7e10a0bb54fc8dbbeff1.

Inherited PR79/PR80 evidence remains unregistered and creates no new dashboard rows. No benchmark family, parser, schema field, CV source, or frontend feature was added.

Validation and review

• Benchmark Frontend CI #355: success, including staleness and Playwright Chromium E2E;
• exact-head project validation run 31101612155: success for docs/static contracts, Python 3.9–3.12 regression, full CPU suite, and sklearn 1.2.2/1.3.2/latest;
• final-head docs validation run 31102102289: success;
• latest fresh review on PR merge: synchronize benchmark dashboard with statgpu 0.2.4 #99 reports no new in-scope finding;
• temporary synchronization/validation workflows are absent from the final tree.

Issue #90 remains open because the final acceptance boundary requires PR #99 to be merged into the PR #76 staging branch and PR #76 to be rechecked for mergeability. No merge was performed without explicit authorization.

[TheHiddenObserver]

Data-completeness clarification

A follow-up review distinguished synchronization integrity from benchmark coverage completeness.

For #90 acceptance:

• all 8 manifest-registered sources exist and parse;
• 0 registered sources are skipped;
• the 1,774-run bundle is deterministic;
• synchronization does not add or fabricate a benchmark family.

Those facts are sufficient for the synchronization scope, but they do not imply complete benchmark coverage.

Follow-up ownership is now explicit:

• benchmark: establish audited source catalog and canonical coverage matrix #100 — audited source catalog, corrected catalog_total/eligible_total semantics, and maintained canonical coverage matrix;
• benchmark: add canonical cross-validation source and dashboard coverage #91 — current structured CV source and dashboard coverage;
• benchmark: complete dashboard production QA and accessibility gate #92 — final production QA/accessibility.

PR #76 and PR #99 descriptions now disclose that 8/8/8 describes the registered bundle only and that the current catalog_total=40 is a raw JSON-artifact count rather than an audited catalog.

[TheHiddenObserver]

Synchronization completed

PR #99 has been merged into the PR #76 staging branch.

Final synchronized state:

master baseline: f95447e8e1bf189f16ccc2d113b3acab2ae53dfe
PR #76 head: 565be3ac9c8b716984ac4a5cb2e8dab5b6be51e9

Acceptance criteria are satisfied:

• PR feat: benchmark dashboard frontend and canonical data pipeline #76 contains the synchronized 0.2.4 master history;
• GitHub reports PR feat: benchmark dashboard frontend and canonical data pipeline #76 mergeable against master;
• no unexplained canonical source-hash or run-identity change occurred;
• the deterministic three-file bundle and deployed assets remain current;
• all post-merge workflows succeeded:
  • Tests #1416;
  • Benchmark Frontend CI #356;
  • Maintenance compatibility #429;
  • Release notes validation #429;
  • Release package validation #435;
• no new benchmark family was introduced by the synchronization;
• all review threads are resolved and no temporary workflow, patch script, diagnostic marker, or generated-asset drift remains;
• PR feat: benchmark dashboard frontend and canonical data pipeline #76 metadata now records the synchronized baseline, final head, current canonical bundle, post-merge validation, known data gaps, and follow-up work.

A final post-merge review on PR #76 found no new synchronization, mergeability, correctness, reproducibility, provenance, documentation, or CI defect.

This issue is complete. PR #76 remains open intentionally for #100 Phase A, #91, and #92 before final integration into master.