⚡ Jev vs Open Model Arena

A side-by-side benchmark dashboard and live leaderboard comparing TypeSafe Jev (System One decision model) against open models (Qwen, DeepSeek, Llama, Gemini Flash, Kimi) via OpenRouter.

---

Features

• 🏆 Dynamic Model Leaderboard: Ranks open models on Fidelity to Jev (% match/calibration), Speed / Latency (ms), Token Consumption, and Dollar Cost.
• 💰 TypeSafe Jev Economics: Visualizes Jev's native rate ($0.042 / 1M input tokens, output tokens free) vs standard LLM token pricing.
• ⚡ Automated Shootout: Hit "Run Full Shootout" to test all curated cheap open models sequentially and populate the leaderboard live.
• 🎯 Multi-Primitive Support: Run Noul (Yes/No probabilities), Choice (categorical options with distributions), and Score (calibrated numeric ratings).
• 🛠️ "Act as Jev" System Prompt Inspector: View and tweak the exact prompt instructing open models to replicate Jev's structured JSON decisions.
• 🔑 Masked API Key Drawer: Keep keys safe with password inputs, visibility toggles, and live configuration status.
• 🚀 Ready for Vercel: Pre-configured with vercel.json for one-click deployment.

---

Quick Start (Local)

# 1. Clone or navigate to the directory
cd jev-compare

# 2. Install dependencies
npm install

# 3. Configure API keys in .env
# TYPESAFE_API_KEY=apikey_...
# OPENROUTER_API_KEY=sk-or-v1-...

# 4. Start the server
npm start

Open http://localhost:3000 in your browser.

---

Deploy to Vercel

See HANDOFF_FOR_CODEX_VERCEL.md for complete Codex & Vercel deployment instructions.

# Using Vercel CLI
npx vercel --prod

Set TYPESAFE_API_KEY and OPENROUTER_API_KEY in your Vercel Project Settings.