How wdobry/laya-playground works
Laya Playground presents recorded examples and three model-driven games, offers an editor for typed questions, and can run locally against Laya’s checkpoints. The main live workflow is a visitor editing a state and questions or playing a game; the browser sends predictions to the loopback server, which validates and routes the request before returning probabilities. Separate branches expose the Laya-versus-Jev benchmark and the agent integration skill. The README documents these capabilities; some browser wiring and server internals were not sampled, so uncertain links are omitted.
Main entry points: server.py Local API server · app.js Landing renderer · run_eval.py Evaluation runner · playground.js Question editor
Components 23
Web experience
- Landing page
index.html - Playground page
playground.html - About page
about.html - Landing renderer
static/app.js - Question editor
static/playground.js - UI helpers
static/ui.js - Result cards
static/cards.js - Recorded examples
static/data
Local inference
- Local API server
server.py - Model routing
poc.py - Laya checkpoints
- Model health
static/health.js
Decision games
- Demo registry
static/demos/index.js - Game simulations
static/demos - Game runtime
static/live.js - Game recordings
static/data
Benchmark
- Benchmark tasks
eval/tasks.json - Dataset builder
eval/build_dataset.py - Evaluation runner
eval/run_eval.py - Benchmark results
static/data/versus.json
Agent guidance
- Integration skill
skills/laya-integration/SKILL.md
Other components
- Visitor
- Jev API
Connections 23
- Visitor → Landing page: opens
- Visitor → Playground page: edits questions
- Visitor → About page: opens
- Landing renderer → Recorded examples: reads examples
- Landing renderer → Result cards: renders answers
- Landing renderer → Integration skill: loads skill
- Question editor → UI helpers: uses prediction
- Question editor → Result cards: renders results
- Question editor → Model health: checks readiness
- UI helpers → Local API server: posts predictions
- Local API server → Model routing: routes request
- Local API server → Laya checkpoints: loads and queries
- Model health → Local API server: reads status
- Game runtime → Demo registry: selects demos
- Demo registry → Game simulations: imports games
- Game runtime → UI helpers: requests predictions
- Game runtime → Game recordings: replays decisions
- Dataset builder → Benchmark tasks: uses task definitions
- Evaluation runner → Benchmark tasks: loads tasks
- Evaluation runner → Local API server: measures Laya
- Evaluation runner → Jev API: measures Jev
- Evaluation runner → Benchmark results: writes report
- Landing renderer → Benchmark results: renders metrics