v1.0 Stable Release

Your Private, Extensible
AI Agent Framework.

Build autonomous agents that live where you work. Lilium connects your IMs (WeChat, Telegram), Calendar, and Email into a unified, privacy-first operating system. Powered by local LLMs or cloud APIs.

Local Privacy First Omni-channel Support TypeScript/Node.js
Lilium AI Dashboard Interface Lilium AI Dashboard Interface

The Future of Work is Agentic

We are on the eve of a paradigm shift in human-computer interaction. In the past, we learned how to operate software; in the future, software will learn how to cooperate with us.

Break Application Silos

Apps are no longer walled gardens. Lilium acts as the glue, connecting WeChat, Email, Calendar, and IM tools into a unified operating system.

Second Brain, Not Search Engine

It doesn't just retrieve information, it understands your context. Evolving from simple Q&A to proactive thinking, planning, and execution.

Data Sovereignty Returns

Your data belongs to you. Lilium prioritizes local models (Ollama, Local Whisper), making privacy no longer the cost of convenience.

Imagine What's Possible

D

Just integrated Lilium into my server alert system. Now whenever CPU spikes, it automatically analyzes logs, sends me a WeChat report, and even tries to restart services. This is #AgenticWorkflow

P

Used to take half an hour to organize meeting minutes. Now Lilium listens to audio, transcribes to text, extracts Action Items, and sends directly to the team group. I don't even open my laptop. 🀯

C

Lilium + Obsidian is a godsend. Notes I jot down in WeChat automatically sync to my knowledge base and get tagged. Finally feel AI is working for me, not me feeding AI.

Why Choose Lilium AI?

Multi-Model Support

Seamlessly switch between OpenAI, Ollama (Local), Claude, etc. Intelligently selects models based on task complexity.

Omni-Channel Access

Support for WeChat, Telegram, Discord, WhatsApp, Lark, DingTalk. Develop once, run everywhere.

Local First & Privacy

Supports running LLMs and speech transcription (Whisper) locally. Keep your data in your own hands.

Plugins & Skills

Built-in web search, file I/O skills. infinitely extend capabilities via a simple JS plugin system.

Voice Interaction

Automatically recognize voice messages, transcribe to text for processing. Supports multi-language TTS.

Long-term Memory

Long-term memory system based on vector database (LanceDB/SQLite), letting AI remember your preferences.

WeChat
Telegram
Discord
WhatsApp
Ollama
OpenAI

Use Cases

πŸ€– Personal Assistant

  • Manage schedules and to-dos
  • Auto-summarize unread messages
  • Voice record inspiration notes

πŸ’¬ Customer Service

  • 7x24h automated inquiry response
  • Accurate Q&A based on knowledge base
  • Auto-transfer complex issues to human

🏠 Smart Home Control

  • Control devices via natural language
  • Family message board
  • Real-time alerts for anomalies

Powerful Modular Architecture

Lilium AI adopts an event-driven micro-kernel architecture, ensuring system stability and scalability.

  • Core Kernel: Handles task scheduling and context management
  • Memory Bus: Efficient vector retrieval and storage
  • Skill Registry: Dynamically load and execute external tools
User Channels (WeChat, TG...)
↓
Lilium Core Agent
Planner Memory Tools
↓
LLM Providers (OpenAI, Ollama...)

Roadmap

v1.0 (Released)

Base architecture, WeChat/Telegram support, OpenAI/Ollama integration, local speech transcription.

v1.2 (In Progress)

Enhanced Web Dashboard, Plugin Marketplace, richer RAG knowledge base support.

v2.0 (Planned)

Multi-modal support (Vision/Image Gen), Distributed Agent Collaboration, Mobile App.

Ready to Build Your Personal AI Assistant?

Open source, free, community-driven. Join the Lilium AI developer community now.

View on GitHub