A cybernetic, real-time benchmark arena that pits frontier large language models (Google Gemini, Mistral, Cohere) head-to-head using LangGraph stateful agent orchestration to evaluate, score, and deliver automated referee verdicts on complex prompts.
- Side-by-Side LLM Duels: Dispatch prompts simultaneously across multiple top-tier AI models (Google Gemini, Mistral AI, Cohere Command).
- LangGraph Referee Engine: An autonomous agentic graph evaluates responses across accuracy, depth, conciseness, and creativity to declare a definitive round winner.
- Cyberpunk Arena Visuals: High-octane dark arena aesthetic featuring neon glass cards, GSAP-powered verdict reveal meters, and animated winner podiums.
- Granular Reasoning Breakdowns: Detailed
ReasoningFooterdisplaying step-by-step scoring metrics, token considerations, and judge critiques. - Rich Output Rendering: Markdown syntax highlighting, code block parsing, table rendering via
react-markdownandremark-gfm.
- Framework: React 19, Vite
- Styling: Tailwind CSS v4 (
@tailwindcss/vite), Custom Cyberpunk Theme - Animations: GSAP 3, Framer Motion 12
- Markdown:
react-markdown,remark-gfm,remark-breaks - Routing & Icons: React Router DOM v7, Lucide React, React Icons
- Runtime: Node.js with TypeScript &
tsx - AI Agent Graph: LangGraph (
@langchain/langgraph), LangChain Core (@langchain/core) - LLM Integrations:
@langchain/google,@langchain/mistralai,@langchain/cohere - Server & Schema: Express v5, Zod
06 - Battle Arena/
├── backend/
│ ├── src/
│ │ ├── ai/ # LangGraph workflow (graph.ai.ts, model.ai.ts)
│ │ ├── config/ # Environment & API configurations
│ │ └── app.ts # Express endpoints
│ ├── server.ts # Server bootstrap
│ └── package.json
└── frontend/
├── public/
│ └── neon-arena-bg.png
├── src/
│ ├── components/ # SolutionCard, VerdictBar, ReasoningFooter
│ ├── pages/ # Arena battle view, loading screen, auth
│ ├── app/App.jsx
│ └── main.jsx
└── package.json
- Node.js (v18+)
- API Keys for Google AI (Gemini), Mistral AI, and Cohere
cd backend
npm installCreate .env in backend/:
PORT=5000
GOOGLE_API_KEY=your_google_gemini_api_key
MISTRAL_API_KEY=your_mistral_api_key
COHERE_API_KEY=your_cohere_api_keyStart server:
npm run devcd ../frontend
npm install
npm run dev| Route | Method | Description |
|---|---|---|
/api/battle |
POST | Submits user prompt to LangGraph duel engine |
/api/models |
GET | Returns available contender LLM configurations |
ISC License. Built for portfolio showcase.