See parent task t-3594 for full context.

What#

Phase 8: BenchmarkRankingOverride.tsx (Debbie)#

• Per-category ranking editor:
  • Show auto-computed ranking (from benchmark scores)
  • Drag & drop to reorder
  • "Lock" toggle per category (prevents auto-updates)
  • Notes field per override
  • Visual indicator: auto rank vs manual rank (e.g., "Auto: #3 → Manual: #1")
• Save override to storage via /api/benchmarks/rankings
• Load existing overrides on mount

Phase 9: BenchmarkStrategyPreview.tsx + strategy bridge (Chris)#

• "Push Rankings to Strategy" button → POST to /api/models/strategy/capability-routing
• Preview panel before push:
  • "If I push these rankings, routing will look like:"
  • Table: category → current model → new model
• Cost projection card:
  • "Switching from $currentModel to $newModel for $category would save $X/month"
  • Based on historical usage + cost data
• Acceptance gate: auto-approve if new scores are >= old scores
• lib/benchmark-strategy-bridge.ts — bridge between benchmark rankings and strategy engine

Deliverables#

• components/Benchmark/BenchmarkRankingOverride.tsx
• components/Benchmark/BenchmarkStrategyPreview.tsx
• lib/benchmark-strategy-bridge.ts
• Ranking override API endpoints

Tags#

[benchmarks, ranking, strategy, integration]