What it does

This is a self-hosted Go server that lets you use ChatGPT, Gemini, and other LLMs inside WeChat without getting your personal account banned. The trick: it tunnels everything through an enterprise WeChat (企业微信) account, which acts as a sanctioned middleman. You get chat, customer service bots, voice messages, image generation, and even a private Milvus vector database for RAG.

The interesting bit

The author is admirably blunt about the architecture’s whole reason for existing — personal WeChat accounts that hook directly into foreign LLM APIs tend to get zapped. Enterprise WeChat is the loophole. The project also doubles as a Dify workflow host, so you can bolt on more complex automation without rebuilding the plumbing.

Key highlights

• Streams responses chunk-by-chunk so you’re not staring at a loading spinner
• Adaptive context window that trims itself, sparing you the manual /reset ritual
• Plugin system with shell, search, and Wikipedia already wired in
• Supports Stable Diffusion 1.5 and OpenAI image generation
• HTTP/SOCKS5 proxy support for the Great Firewall-afflicted
• Docker Compose deployment with Postgres + Redis + Milvus

Caveats

• v1.0.0 is in active development; the README steers you back to v0.6.6 for stability
• Database migration to Postgres in v1 means existing Redis-only setups need care
• Several promised features (Midjourney API, web admin UI, long-term memory) are still unchecked on the roadmap

Verdict

Worth a look if you need LLM access inside China’s WeChat ecosystem and can stomach enterprise WeChat’s registration hurdles. Skip it if you just want a quick personal chatbot — the setup is decidedly not quick.