How hanggaa/benchmark works
The repository combines an AI-assisted MVP workflow (research through build) with a benchmark suite and its results dashboard. The documented workflow is represented from its prompt/template artifacts; the sampled frontend shows dashboard composition and local-runner dispatch instructions. Benchmark runner internals were not sampled, so the graph does not infer their internal call chain.
Main entry points: App.tsx Dashboard App · types.ts Benchmark Types · LeaderboardTable.tsx Leaderboard · CaseInspectorModal.tsx Case Inspector
Components 18
MVP Workflow
- Idea Research
part1-deepresearch.md - MVP Requirements
part2-prd-mvp.md - Technical Design
part3-tech-design-mvp.md
Project Artifacts
- Agent Instructions
templates - Tool Adapters
templates/tool-adapters
Benchmark Suite
- Benchmark Runner
benchmarks - Benchmark Results
src/data/benchmark-data.json
Results Dashboard
- Dashboard App
src/App.tsx - Leaderboard
src/components/LeaderboardTable.tsx - Case Inspector
src/components/CaseInspectorModal.tsx - Top Model Card
src/components/HeroRankCard.tsx - Cost Accuracy Chart
src/components/ScatterChart.tsx - Benchmark Categories
src/components/CategoryCards.tsx - CLI Quick Start
src/components/CLIQuickStart.tsx - Benchmark Types
src/types.ts
Other components
- Product Builder
- AI IDE
- Working MVP
Connections 22
- Product Builder → Idea Research: frames idea
README.md - Idea Research → MVP Requirements: informs requirements
README.md - MVP Requirements → Technical Design: defines scope
README.md - Technical Design → Agent Instructions: guides setup
README.md - Agent Instructions → AI IDE: provides context
README.md - Tool Adapters → AI IDE: adapts instructions
README.md - AI IDE → Working MVP: builds
README.md - Benchmark Runner → Benchmark Results: records runs
- Benchmark Results → Dashboard App: loads data
src/App.tsx - Dashboard App → Leaderboard: renders
src/App.tsx - Dashboard App → Top Model Card: renders
src/App.tsx - Dashboard App → Cost Accuracy Chart: renders
src/App.tsx - Dashboard App → Benchmark Categories: renders
src/App.tsx - Dashboard App → CLI Quick Start: renders
src/App.tsx - Dashboard App → Case Inspector: opens selected case
src/App.tsx - CLI Quick Start → Benchmark Runner: shows command
src/components/CLIQuickStart.tsx - Leaderboard → Case Inspector: selects case
- Dashboard App → Benchmark Types: uses data types
src/App.tsx - Case Inspector → Benchmark Types: reads item fields
src/components/CaseInspectorModal.tsx - Leaderboard → Benchmark Types: uses summaries
src/components/LeaderboardTable.tsx - Top Model Card → Benchmark Types: uses summary
src/App.tsx - Cost Accuracy Chart → Benchmark Types: uses summaries
src/components/ScatterChart.tsx