What it does

This is glue code with ambition: it wires the OpenAI API into WeChat via wechaty, so your personal account can field ChatGPT conversations, DALL·E image requests, and even Whisper transcriptions. You scan a QR code to log in, set a trigger keyword, and suddenly your chat app has a co-pilot.

The interesting bit

The deployment ergonomics are surprisingly thorough for a weekend project — Railway one-click deploy, Fly.io template, Docker Compose, plus a companion Cloudflare Workers proxy for regions where OpenAI is cranky about direct access. That 13K stars didn’t come from nothing; it came from “just works” hosting instructions.

Key highlights

• Supports GPT-3.5-turbo, DALL·E, and Whisper through a single WeChat bot interface
• Conversation memory and slash commands (/cmd prompt, /cmd clear) for session management
• Keyword triggers and block-word filters for both private and group chats
• Multiple deployment paths: Railway, Fly.io, Docker, Docker Compose, or raw Node.js 18+
• Companion openai-proxy project for API endpoint customization

Caveats

• Archived. The maintainer shut it down; no updates, no issue triage, likely bit-rot against current WeChat web protocol changes
• Model support is frozen at gpt-3.5-turbo and gpt-3.5-turbo-0301; no GPT-4, no function calling
• Requires a WeChat account that can still log in via web WeChat, which increasingly excludes newer accounts

Verdict

Worth studying if you’re building a similar bridge — the deployment patterns and wechaty integration are solid reference material. Skip it for production use unless you plan to fork and maintain the protocol layer yourself.