A WeChat bot that sidesteps the ban hammer

What it does

This is a Python bot that plugs ChatGPT (and a dozen other LLMs via litellm) into WeChat. It handles private chats, group chats, multi-turn conversations with memory, and DALL-E image generation. You configure triggers, prompts, and session behavior in a JSON file.

The interesting bit

Instead of using the usual itchat or wechaty web-protocol routes that get accounts flagged, it injects a DLL into a specific old Windows WeChat client (3.2.1.121) and reads messages through a Windows hook. The author calls this “no risk”; what it actually means is you’re trading one risk profile for another—specifically, running a pinned, EOL WeChat build with a memory injection tool.

Key highlights

• Supports GPT-3.5, GPT-4, Claude, PaLM, and anything else litellm covers
• Plugin system with an extension interface; one example plugin fetches TikTok content
• Session memory with configurable expiration and token limits
• Image generation triggered by prefixes like “draw” or “imagine”
• Windows-only; Linux support is a distant “probably” contingent on a Docker sandbox project

Caveats

• Requires an exact, outdated WeChat version and a third-party DLL injector—neither of which WeChat’s owner sanctions
• The “no risk” claim is relative; you’re still modifying a closed-source client at runtime
• Python 3.13+ required, which is unusually aggressive for a project with ~600 stars

Verdict

Worth a look if you need a WeChat-LLM bridge and have accepted the DLL-injection tradeoffs. Skip it if you wanted something cloud-native, cross-platform, or that doesn’t involve pinning a 2021-era chat client.