🎥 Watch the Demo

🧩 What This App Does

This dashboard lets you watch (or join!) teams of Large Language Models (LLMs) play Codenames against each other. Two teams — Red and Blue — face off in a 4v4 format. Each team has a Boss and three Agents working together to identify their team's words before the other side does.

🤖 How It Works

• LLM Teams: You can assemble teams using different LLMs (e.g., GPT, Claude, Gemini, or OpenSource models...).
• Human Mode: You can also jump in as a Boss yourself, giving clues to your AI teammates and seeing how well they interpret your hints.
• Observation Mode: Prefer to just watch? Sit back and enjoy the game unfold, analyzing how different models reason, cooperate, and sometimes hilariously misfire.

🧠 Why It's Interesting

• Compare LLM reasoning styles: See how different models interpret subtle associations and language cues.
• Team Dynamics: Watch how collaboration (or confusion) emerges between AIs when they have to coordinate across multiple turns.
• Human-AI Interaction: Experiment with leading a team of LLMs and discover how clearly (or creatively) you need to communicate to win.
• Benchmarking & Analytics: All games are stored in a database. The Stats section includes model win/loss rates, performance comparisons between model families and leaderboards

🕹️ Main Features

• Create and customize teams with any available LLMs.
• Switch between AI vs AI and Human&AI vs AI modes.
• View reasoning and chat logs for each model's decisions.