What it does

Kirara AI is a multi-platform chatbot framework that connects mainstream LLMs to Chinese and global messaging apps. You run it as a Docker container or Python service, configure models and platform adapters in a WebUI, and it handles group chats, private messages, image generation, and text-to-speech replies. It also exposes HTTP APIs for custom integrations.

The interesting bit

The project treats “persona” as a first-class feature, not an afterthought. It ships with preset character files (the README highlights a “catgirl” mode) and supports user-shared prompt libraries, turning the usual “ChatGPT wrapper” into something closer to a roleplay engine with workflow automation.

Key highlights

• Supports OpenAI, DeepSeek, Claude, Gemini, Qwen, Mistral, Kimi, and Ollama backends
• Native adapters for QQ, WeChat, Telegram, enterprise WeChat, and public WeChat accounts; Discord and Feishu marked “restructuring”
• Built-in Stable Diffusion, Flux, and Midjourney image generation
• Workflow system with visual editor, plugin marketplace, and conditional triggers
• Voice replies via Azure TTS or VITS; multi-account and multi-round conversation support
• Includes frpc for tunneling out of restricted networks

Caveats

• Discord and Feishu support is currently “restructuring” — feature parity unclear
• README is entirely in Chinese; documentation and community support assume Chinese fluency
• Windows voice features require manual VC++ redistributable installation

Verdict

Worth a look if you need a self-hosted bot bridging Chinese social platforms to modern LLMs, especially with character/persona requirements. Skip if you need polished Discord integration or don’t read Chinese.