SMNETSTUDIO/WeChat-AI

What it solves

It provides a self-hosted framework for creating AI-powered roleplay chatbots on WeChat. It bridges the gap between Large Language Models (LLMs) and the WeChat ecosystem, allowing users to deploy bots with specific personas, memories, and the ability to interact with real users via a managed service.

How it works

The system connects to Tencent's iLink to handle WeChat messaging. It uses a multi-node architecture where several identical server instances share a remote Redis database for storing bot tokens, personas, and stickers. Authentication is handled via LINUX DO OAuth. For AI capabilities, it integrates with OpenAI-compatible LLMs and can use a separate AI gateway (deployed on Hugging Face) for web search and custom model access. It also includes a visual orchestration tool called Chatflow for designing bot behaviors.

Who it’s for

Users who want to host their own AI roleplay bots on WeChat, developers interested in integrating LLMs with messaging platforms, and community managers who need an admin dashboard to monitor token usage and moderate content.

Highlights

• Visual Orchestration: Includes a Chatflow editor for designing complex bot interactions beyond simple prompts.
• Multi-Node Deployment: Supports homogeneous deployment with Cloudflare Worker load balancing for stability.
• Multimodal Capabilities: Supports inbound image understanding (vision) and built-in voice transcription.
• Persona Management: Features a user center for assigning personas and managing approved chat peers.
• Rich Media Support: Integrates a sticker library that allows the AI to reference and send specific image stickers via iLink CDN.