Lore

Open-source AI Life Simulation — AI characters live autonomously in a virtual world. You're part of it, not the center of it.

English | 中文

What is Lore?

Lore is not a chatbot. It's a living world powered by AI agents.

Each agent has their own life — career, relationships, personality, and memories. They go to work, fall in love, make friends, and experience life events. The world keeps running whether you're online or not.

You're not the center of their world. You're part of it.

Key Features

• Autonomous AI Agents — Every agent thinks with LLM. No scripted behavior. The delivery guy might quit and start a company
• Living World — The world runs 24/7. World Agent handles natural disasters, economy, social changes
• Two World Types — Random mode (set age, location, explore) or History mode (time-travel into a historical figure)
• Deep Interaction — Chat, upload photos/videos, build relationships, get rejected, get flirted with
• Virtual Platforms — Agents use simulated YouTube, TikTok, Twitter. Post selfies, browse content, react to your posts
• Unrestricted Agents — Agents have complete lives. They can start businesses, compete with you, fall in love, disappear
• Event Cards — Life events pop up as cards. Choose to intervene or watch from the sidelines
• God Mode — Observe every agent's thought process, trigger world events, see the full picture
• Community Presets — Historical eras, sci-fi worlds, custom scenarios — written in YAML, shared by the community
• Privacy First — All data stored locally (SQLite). No cloud dependency
• Multi-LLM — Works with OpenAI, Claude, DeepSeek, Kimi, local models, and more

Quick Start

# Install
npm install -g lore

# Run
lore

First run will:

1. Create ~/.lore/ data directory
2. Initialize SQLite database
3. Start the world engine
4. Open your browser → http://localhost:3952

Development

# Clone
git clone https://github.com/ksxh0524/lore.git
cd lore

# Install dependencies
pnpm install

# Start development (server + client in parallel)
pnpm dev

Tech Stack

• Frontend: React 19 + TypeScript + Vite + shadcn/ui + Tailwind CSS + zustand
• Backend: Node.js + Fastify + WebSocket
• Database: SQLite + Drizzle ORM + vec0 (vector search)
• AI: Vercel AI SDK + OpenAI-compatible API (DeepSeek, Kimi, Qwen, Claude, Gemini...)
• Package: pnpm monorepo

Project Structure

lore/
├── packages/
│   ├── server/       # Backend + World Engine + Agent System
│   ├── client/       # Frontend React PWA
│   └── shared/       # Shared types
├── docs/             # Technical documentation
├── AGENTS.md         # AI Agent coding guide
├── CONTRIBUTING.md   # Contribution guide
└── README.md

See Technical Documentation for architecture details.

Roadmap

• Phase 0: Project scaffolding + documentation
• Phase 1: World initialization + single agent + basic sandbox + chat + basic economy (MVP)
  • Project setup with pnpm monorepo
  • Database layer (SQLite + Drizzle ORM)
  • REST API and WebSocket
  • LLM provider abstraction
  • Agent Runtime with memory system
  • World initialization system
  • Tick scheduler and world engine
  • Basic economy system
  • Frontend UI with React
• Phase 2: Memory engine + multi-agent + relationships + virtual platforms + image generation
• Phase 3: World Agent + autonomous behavior + god mode + push notifications
• Phase 4: History mode + community presets + advanced sandbox + factions
• Phase 5: Full economy + multi-modal + PWA + npm publish
• Phase 6+: Multi-user worlds + community ecosystem + 10k+ agents

Contributing

Contributions are welcome! Please read CONTRIBUTING.md first.

License

MIT