[pelikhan]

Summary

Optimize only the Repository overview (repository-detail) by declaring query-binding: page. Its ordinary widgets now share one abort-scoped canonical subscription instead of independently recomputing repository-activity. Other pages keep their existing independent subscriptions.

The production view-payload compiler shares equivalent scoped dependency graphs within one request while retaining distinct view aliases. Per-view filters, search, ordering, limits, source quality, and provenance remain intact; there is no cross-request cache. Database updates still refresh the active page, and navigation or context changes dispose the old subscription. Page-bound failures render an explicit unavailable state.

Measured result

Fully settled default-page readiness fell from a 26.59 s median to 5.88 s: 77.9% less latency / 4.52× faster.

Pair	Order	Before	After
1	Before → After	24.10 s	5.88 s
2	After → Before	27.61 s	5.83 s
3	Before → After	26.59 s	6.43 s

Each before run started five page subscriptions (four duplicate activity widgets and page chrome); each after run started one. All active widget rows and evidence availability, completeness, freshness, and timestamps matched their before results.

Method: the unchanged base checkout at 2de03249512f65fc0eb7ca2a438eac647205f2e2 and the optimized build were alternated on the same Chromium origin/database and pinned public manifest. Each round forced a full app reload, waited for eager ingestion to finish, then navigated to #page-repository-detail?repository=githubnext%2Fgh-aw-cao. Readiness waits for all active page requests and visible views, excluding ingestion and the 400 ms quiet-period confirmation. No CPU/network throttling; 1440×1000 viewport. Alternating pair order and three samples account for shared-machine variability.

Every round verified the same 575,582 canonical records, including 4,073 runs, 695 workflows, and 6 repositories. This reproduces the reported deployment's fourfold repeated dependency evaluation without treating its single 27.86 s sample or the abandoned diagnostic run as the paired baseline.

Pinned manifest SHA-256: cce19a1f7237786b47a363678ea51e6a685e93a2eca21dd1bd80aac82b605808.

Coverage

Regression tests require a single shared import-evidence dependency graph and less than 40% of the independent row-operation budget. Scoped values and partial/stale provenance are compared across time horizons, repository/organization/workflow/mode filters, and search. Additional tests cover distinct filters/limits, binding validation, page-lifetime aborts, refreshes, query errors, and browser-worker updates to metrics and health in Table mode and mobile layout.

Passed:

• 408 focused unit tests, including all new regressions and an isolated retry of the original full-suite timeout.
• Dashboard lint, strict typecheck, and document validation; root lint.
• Production-worker browser regression for metrics, refresh, Table-mode health, and mobile layout.
• Canonical data-worker heap-budget guard against deployed data.
• Lighthouse CFO/CTO/CSO regression scenarios against the pinned data: 12 desktop/mobile × empty/full cold-navigation audits, one repeat each, all budgets passed; scores 97–100. The repeated paired page measurements above remain the primary optimization evidence.
• Workflow compilation: 71 workflows, with two existing warnings.
• Documentation build and static agent-discovery validation.
• Agent-catalog regeneration produced no changes.

Full-suite caveats, not hidden or loosened:

• The first full dashboard unit run passed 3,604 tests, skipped one, and timed out in one unchanged campaign-query test. A bounded-worker full retry passed 3,602 but timed out in three unchanged storage/native-aggregate tests that passed the first run. The original campaign-query suite passes in the final 408-test focused run. Shared-machine contention makes the full-suite timing gates non-clean.
• Full browser run: 291 passed, one skipped, two failures. The Safari offline-worker timeout passes all five cases when isolated. The other failure is the existing core-text size budget: unchanged baseline 78,039 bytes already exceeds 76,500; this change yields 78,062 (+23). No unrelated page or size-budget changes are included.
• The optional full root check was stopped after an unrelated installer test failed; that exact installer test passes in isolation. Required focused checks, workflow compilation, and documentation build completed separately.

Populated desktop and mobile screenshots are attached in a PR comment.

[pelikhan]

Populated Repository overview after the optimization, using the same fully ingested pinned dataset as the paired measurements. Desktop: 1440×1000. Mobile: 390×844.

[github-actions]

Dashboard dead views

Dead element views (0):
Dead dashboard pages (0):
Analyzed 8 element renderers and 66 dashboard pages across 1 dashboard.json file(s) from dashboard/site/src/components/ui-elements.js.

⚠️ This reachability report is informational and does not block the pull request.
🔎 View the workflow run and report artifact.

[Copilot]

🟡 Changes recommended

Page binding can omit an unscoped metric’s source when another metric shares that source with global filters.

1 open issue

• Shared derived source omitted for unfiltered metric under page binding · New

What changed in this PR

Optimizes Repository overview by sharing one page-level subscription and equivalent query dependencies within each request.

Changes:

• Adds opt-in query-binding: page with validation and documentation.
• Shares scoped dependency graphs while retaining separate view aliases.
• Adds regression coverage for performance, refreshes, cleanup, and failures.

File	Description
docs/​dashboard-language-specification.md	Defines page-binding behavior.
dashboard/​site/​test/​unit/​view-payload-compiler.test.js	Tests shared graphs and distinct limits.
dashboard/​site/​test/​unit/​validator.test.js	Tests binding validation.
dashboard/​site/​test/​unit/​repository-activity-query-optimization.test.js	Checks scope, quality, and operation budgets.
dashboard/​site/​test/​unit/​presenter.test.js	Tests subscription lifecycle and failures.
dashboard/​site/​test/​fixtures/​repository-overview-contract.json	Defines overview contract expectations.
dashboard/​site/​test/​e2e/​repository-overview.spec.js	Checks worker refreshes and responsive rendering.
dashboard/​site/​src/​validator-pages.js	Validates binding values.
dashboard/​site/​src/​specification.js	Allows the custom-page binding key.
dashboard/​site/​src/​presenter.js	Implements page-level binding and failure states.
dashboard/​site/​src/​data/​queries/​view-payload-compiler.js	Reuses equivalent scoped dependency graphs.
dashboard/​site/​dashboard-fragments/​repositories.json	Enables binding for Repository overview.

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Campaigns page binding bypasses its early inventory preview, delaying identities and navigation links until full activity ingestion.

1 open issue

• Page binding bypasses Campaigns partial-source previews · New

1 resolved since last review

• Shared derived source omitted for unfiltered metric under page binding

_{Review effort: Balanced}

[Copilot]

🟡 Changes recommended

Continuation loading rejects valid tokens, and backend-ineligible previews can trigger complete queries before ingestion is ready.

2 open issues

• Preserve compilation scope for continuation loading · New
• Filter preview discovery by active backend · New

1 resolved since last review

• Page binding bypasses Campaigns partial-source previews

_{Review effort: Balanced}

[pelikhan]

Superseded by the single consolidated draft PR #19015 at f0b769c. The reviewed source changes and preservation corrections are collected there, with regenerated catalogs and combined validation. Closing without merging or deleting the source branch; history and original evidence remain available.