yenche123/liubai

TL;DR

Liubai is an open‑source, AI‑enhanced personal knowledge‑base / task‑manager that works across web, WeChat, and VS Code. It lets you capture notes, schedules, to‑do items, and attach rich media, then optionally query or augment them with large language models (DeepSeek, Kimi, MiniMax, etc.) directly from WeChat.

What it does

Feature	What you get
Atomic cards	Every piece of information (text, image, file, tag, status) lives on its own “card” that can be expanded or hidden.
Cross‑platform UI	Responsive web app (PWA), WeChat mini‑program, and a VS Code extension – all share the same backend.
AI assistance	Connects to several LLM providers; you can ask the AI to summarise, generate, or draw based on the content of a card, all from within WeChat.
Offline‑first	The web front‑end uses a service‑worker + Dexie (IndexedDB) so you can read/write cards without an internet connection; changes sync when you’re back online.
Privacy controls	Each card can be marked “local only” or “do not let AI read”, giving you fine‑grained data‑ownership.
Sync & collaboration	Cloud sync via the Laf backend; collaborative task lists that can be shared with WeChat contacts.
Dark / light mode & i18n	Built‑in theme switch and Chinese/English language toggle.
VS Code quick‑note	A lightweight extension to create cards without leaving the editor.

How it works (high‑level architecture)

liubai-backends/
│   └─ liubai‑laf   ← backend built on the Laf framework (cloud‑functions live here)

liubai-frontends/
│   ├─ liubai‑web          ← Vue 3 + Vite PWA (offline storage via Dexie)
│   ├─ liubai‑weixin       ← WeChat mini‑program (Skyline + Chat Tool API)
│   └─ liubai‑vscode‑extension ← VS Code plugin (TypeScript, bundled with esbuild)

liubai‑docs/ ← documentation site built with VitePress

The backend exposes APIs for CRUD on cards, sync, and routing AI requests to the configured LLM services.
The web PWA talks to the backend, caches data locally, and registers a service‑worker for offline use.
The WeChat mini‑program uses the official 客服消息 API and the new Chat Tool API to let users chat with AI assistants and manage tasks directly in WeChat.
The VS Code extension provides a command palette entry that creates a new card and pushes it to the same backend, so notes stay in sync with the other clients.

Who might find it useful?

Knowledge workers who want a single place for notes, calendar events, and to‑do items that can be enriched by LLMs.
WeChat‑centric users (especially in China) who prefer doing everything inside the app they already use.
Developers who like an open‑source, self‑hostable alternative to proprietary “AI‑powered” note‑taking services.
Privacy‑conscious people because each card can be kept local and excluded from AI processing.

Getting started

Read the docs – https://docs.liubai.cc (install guide, API keys for LLMs, deployment steps).
Deploy the backend – follow the Laf instructions; the repo already contains a cloud-functions/ folder.
Run the web client – cd liubai-frontends/liubai-web && bun install && bun run dev.
Optional – add the VS Code extension from the marketplace or build it locally, and/or scan the QR code to use the WeChat mini‑program.

License

AGPL‑v3 – you can run, modify, and host your own instance, but any network‑served modifications must be made public.

All details above are taken directly from the repository’s README.