ai-town
A MIT-licensed, deployable starter kit for building and customizing your own version of AI town - a virtual town where AI characters live, chat and socialize.
open-sourceagent-frameworks
9.6k
Stars
+180
Stars/month
0
Releases (6m)
Star Growth
+35 (0.4%)
Overview
AI Town 是一个基于 MIT 许可证的可部署启动套件，用于构建虚拟城镇，其中 AI 角色可以生活、聊天和社交。该项目受到「生成式代理：人类行为的交互式模拟」研究论文的启发，提供了强大的技术基础，支持共享全局状态、事务处理和仿真引擎。与大多数基于 Python 的仿真器不同，AI Town 使用 JavaScript/TypeScript 构建，技术栈包括 Convex（游戏引擎、数据库、向量搜索）、可选的 Clerk 认证、PixiJS 渲染引擎。默认使用 llama3 模型和 mxbai-embed-large 嵌入，支持本地推理（Ollama）和云端 LLM（OpenAI、Together.ai）。该项目旨在从简单的实验项目扩展到多人游戏，为 AI 社交仿真提供了一个可定制的平台。
Deep Analysis
Key Differentiator
The only JavaScript/TypeScript-based AI town simulation with built-in infrastructure (Convex backend, real-time state, pixel rendering) — filling the gap where most agent simulators are Python-only and lack deployment-ready architecture
⚡ Capabilities
- • Deployable starter kit for virtual towns with autonomous AI characters that interact and socialize
- • Real-time simulation engine with shared global state, transactions, and pixel art rendering
- • Multi-agent interaction with memory systems and contextual decision-making
- • Background music generation and customizable environments via Tiled Map Editor
- • Built-in infrastructure via Convex backend with dashboard for data browsing
🔗 Integrations
Ollama (default LLM)OpenAITogether.aiAnthropicGroqConvex (backend/database)Clerk (auth)Replicate (art/music generation)PixiJS (rendering)Vercel
✓ Best For
- ✓ Developers building interactive AI character simulations or virtual worlds with JS/TS
- ✓ Researchers exploring emergent multi-agent behavior in simulated environments
✗ Not Ideal For
- ✗ Production-ready game engines — use Unity or Unreal with AI plugins for commercial games
- ✗ Python-first AI researchers — use generative_agents (Stanford) for Python-based simulation
Languages
TypeScriptJavaScript
Deployment
Convex Cloud (free tier)Docker Compose self-hostedFly.ioVercel (frontend)Pinokio one-click installerLocal development
Pricing Detail
Free: Free to start — Convex free tier with auto-scaling
Paid: Scales with Convex pricing as usage grows
⚠ Known Limitations
- ⚠ Simulation pauses after 5 minutes of browser inactivity
- ⚠ Changing LLM providers requires database wipe and restart
- ⚠ Windows users need WSL2 setup
- ⚠ Designed as a starter kit — production use requires significant customization
Pros
- + 强大的技术架构，基于 Convex 提供共享状态、事务处理和仿真引擎支持
- + 高度可配置，支持多种 LLM 提供商（本地 Ollama、OpenAI API、Together.ai）
- + 活跃的开源社区，拥有 9,622 个 GitHub 星标和 Discord 社区支持
Cons
- - 设置复杂，需要配置多个服务（Convex 后端、LLM 提供商、可选认证）
- - 运行多个 AI 代理会消耗大量计算资源
- - 仍处于实验性质，适合研究和探索而非生产环境
Use Cases
- • 研究 AI 代理行为和社交动态的学术项目
- • 构建多人 AI 驱动的仿真游戏
- • 探索 AI 社交互动的教育项目
Getting Started
1. 克隆仓库并安装依赖项 2. 设置 Convex 后端并配置 LLM 提供商（Ollama/OpenAI/Together.ai） 3. 本地部署或生产部署，配置 AI 角色参数