wtf is wechat-ai?

WeChat-AI is a self-hosted service that lets you run an AI chatbot inside WeChat, Tencent's popular Chinese messaging app, so it can hold roleplay conversations with users. It connects directly to Tencent's official iLink interface for WeChat bots, stores its data in a remote Redis database, and handles login through LINUX DO, a community OAuth provider. Once set up, users log in with their LINUX DO account and use a user center to scan a QR code and connect their own WeChat bot account, approve which people are allowed to chat with it, and assign it a persona to roleplay as. To start a conversation, a bound user mentions another user by name inside WeChat, and once that person replies with an agree command, messages are relayed both ways between them through the bot. Replies can include both text and images pulled from a sticker library, and a typing indicator shows while a response is being generated. There is an admin dashboard for managing the whole deployment: viewing statistics, tracking how many AI tokens are being used each day, managing users and bots, reviewing submitted stickers for anything malicious, and overseeing multiple deployed nodes. The chatbot connects to any OpenAI-compatible AI service for its responses, and users can configure custom models and enable web search, which is routed through a separate tools gateway so the main server never talks to outside APIs directly. There is also a visual Chatflow editor for building more structured conversation logic beyond a simple prompt. The system is built to run across multiple servers sharing one Redis database, with a Cloudflare Worker handling health checks and routing traffic between them, and it supports incremental updates that restart automatically. Setup involves installing dependencies with pnpm, filling in a Redis URL, an AI provider key, and LINUX DO OAuth credentials, then running it locally or deploying it with Docker. The project explicitly warns that using Tencent's WeChat bot capabilities carries a risk of the account being rate-limited or shut down, and that roleplay conversations are sent to outside LLM APIs, so users should consider their own privacy needs. It is licensed under Apache-2.0 with the Commons Clause added, which permits personal and non-commercial use only and forbids charging for hosting, offering paid support, or reselling the code.