How pman369/ai-battle-royale-developer-arena works
This Paperclip plugin runs competitive, budget-constrained rounds in which agents submit code patches for isolated evaluation. The match engine tracks spending, grades submissions, and advances or ends rounds; supporting capabilities include governance audits, durable event replay, comparative metrics, dataset export, and a dashboard. The README documents the main workflow and dashboard features. Some worker integrations are named in the source index but their call sites are not shown, so those links are marked unverified rather than asserted.
Main entry points: worker.ts Match engine · types.ts Arena data model · eval_metrics.ts Evaluation metrics · routine.ts Routine coordinator
Components 15
Match orchestration
- Routine coordinator
src/routine.ts - Match engine
src/worker.ts
Patch evaluation
- Sandbox manager
src/sandbox.ts - Governance auditor
src/governance.ts
Audit and insights
- Evaluation metrics
src/eval_metrics.ts - Activity logger
src/activity_log.ts - JSONL event files
- Dataset exporter
src/dataset_exporter.ts - Evaluation dataset
Paperclip plugin
- Arena cockpit
src/ui/index.tsx - Plugin manifest
src/manifest.ts - Arena data model
src/types.ts
Other components
- Paperclip host
- Contestant agents
- Target test command
Connections 21
- Paperclip host → Plugin manifest: loads plugin
src/manifest.ts - Paperclip host → Routine coordinator: schedules ticks
src/routine.ts - Routine coordinator → Match engine: advances rounds
src/routine.ts - Contestant agents → Match engine: reports spend
README.md - Match engine → Sandbox manager: evaluates patches
src/worker.ts - Sandbox manager → Target test command: runs tests
src/sandbox.ts - Match engine → Governance auditor: audits patches
src/worker.ts - Match engine → Activity logger: persists events
src/worker.ts - Activity logger → JSONL event files: appends and reads
src/activity_log.ts - Evaluation metrics → Arena cockpit: supplies scorecard
README.md - Match engine → Evaluation metrics: computes metrics
src/worker.ts - Match engine → Dataset exporter: generates dataset
src/worker.ts - Dataset exporter → Evaluation dataset: exports JSONL
README.md - Paperclip host → Arena cockpit: hosts dashboard
- Match engine → Arena data model: uses match model
src/worker.ts - Routine coordinator → Arena data model: uses schedule config
src/routine.ts - Sandbox manager → Governance auditor: shares patch submission
src/governance.ts - Evaluation metrics → Arena data model: reads events and agents
src/eval_metrics.ts - Activity logger → Arena data model: stores typed events
src/activity_log.ts - Dataset exporter → Evaluation metrics: accepts summary
src/dataset_exporter.ts - Arena cockpit → Arena data model: displays match state
src/ui/index.tsx