Skip to content

PR Flow

What it answers

  • Where work stalls (pickup, review, merge)?
  • Is the bottleneck capacity or latency?

Implementation

PR Flow is implemented through multiple complementary visualizations:

1. Flame Diagram (Single PR Timeline)

Location: /prs/[pr_id]/page.tsx → FlameDiagram component

Shows the lifecycle of a single PR with: - States: active, waiting, blocked, ci - Categories: planned, unplanned, rework - Timeline visualization with color-coded segments - Duration breakdown by state

Use case: Diagnose why a specific PR took long to merge.

2. State Flow Sankey

Location: /work → Flow tab → State Flow sub-tab

Shows work item state transitions across the team: - Source nodes: initial states - Target nodes: terminal states - Link width: volume of transitions - Identifies common flow paths and bottlenecks

Use case: See where work commonly gets stuck across all items.

3. Review Heatmap

Location: /work → Heatmap tab

Shows review wait patterns by time: - X-axis: time of day - Y-axis: day of week - Cell intensity: review wait density

Use case: Identify when review latency is highest.

Backing computations

  • Stage timestamp extraction per provider (pr_created_at, first_review_at, merged_at)
  • Windowed aggregation by scope
  • State duration calculations in work_item_state_durations_daily

API Endpoints

  • GET /api/v1/flame?entity_type=pr&entity_id={id} - Single PR timeline
  • POST /api/v1/sankey with mode=state - State transitions
  • GET /api/v1/heatmap?type=temporal_load&metric=review_wait_density - Review patterns