HYDI-40: Filter dashboard items by Project

[oz-by-warp]

Implements HYDI-40

Changes

• Added a dashboard project dropdown filter (All Projects + discovered project keys from dispatch runs).
• Added server-side filtering by project query parameter so only runs from the selected project are rendered.
• Preserved filter state across the existing Hide Done toggle.
• Added a Project column to the dashboard table for clarity.
• Added route test coverage for project dropdown rendering and project-specific row filtering.
• Updated dashboard docs to include the new project filter behavior.

Validation

• npm test
• npm run test:coverage
• npm run typecheck

UI verification

This ticket touches dashboard UI rendering. Automated screenshot/snapshot iteration via Playwright MCP was not executed in this environment because Playwright MCP/browser control tools are not available in this run.

Conversation: https://app.warp.dev/conversation/657f393a-f151-4dcc-a24c-a92416acc122
Run: https://oz.warp.dev/runs/019e4ff5-2b36-7145-b07f-8b5ddba57893

Co-Authored-By: Oz oz-agent@warp.dev
This PR was generated with Oz.

[github-actions]

PR Review — Principal Engineer Assessment

Verdict: Ready with fixes
Risk: Medium
Scope: 63dcad700026370a7a94a5cba8671ab37c4fc7ff..661db629cdacc002f0e1c651d729cc5726e4ccf0
Requirements: Jira HYDI-40 (Filter dashboard items by Project)

Strengths

• Project filter is implemented as a cohesive route-level slice (query parsing, options derivation, filtering, rendering).
• State preservation between project selection and hide-done toggle is handled cleanly via query params.
• Added route test covers dropdown rendering and project-specific row filtering.
• Dashboard docs were updated to reflect the new UI behavior.

Architecture Assessment

• Vertical slice quality: Strong — route change includes UI, filtering logic, tests, and docs in one coherent slice.
• Deep module quality: Mixed — rendering and aggregation logic remain centralized, but filtered-row behavior now diverges from summary counts, reducing conceptual cohesion.
• Testability/reasoning quality: Mixed — row filtering is tested, but count semantics under project filtering are not validated.

Third-Party Contract Verification

• No third-party API contract surface was changed in this PR.
• Jira client usage remains on existing getIssue(..., ["status"]) path with unchanged request/response assumptions.

Unmapped Findings

(findings whose lines are not in the diff — inline comment not possible)

• None.

Action Plan For Implementing Agent

1. [REV-001] Recompute dashboard summary counts from the already-filtered run set (or introduce a query-aware count source) so stat cards match the selected project/hide-done filters.

actions:
  - id: REV-001
    severity: important
    location: src/routes/dashboard.ts:143
    change_required: Replace global `countRows`-based stat card values with counts derived from the active filtered set (`projectFilteredRuns` and hide-done state) or fetch counts with equivalent query constraints.
    acceptance_check: With mixed-project fixtures, requesting `/dashboard?project=HYDI` shows stat cards whose totals only reflect HYDI rows; with `hideDone=1`, done-category items are excluded from both table rows and stats.

[github-actions]

[REV-001] Important — Stats cards ignore active project filter

Problem: The table rows are filtered by selectedProject, but summary stat cards are still populated from global countRows (all projects), so counts can disagree with what the table is showing.
Impact: Users can make incorrect operational decisions because dashboard totals do not represent the currently selected project view.
Required fix: Compute stats from the same filtered dataset used for rendering rows (or fetch query-aware counts keyed by project and hide-done state).
Acceptance check: For a dataset with multiple projects, /dashboard?project=HYDI must display running/queued/blocked/succeeded/failed/stale counts that match only HYDI runs; adding hideDone=1 must also remove done-category tickets from both rows and counts.