ChatLunaLab/chatluna

多平台模型接入，可扩展，多种输出格式，提供大语言模型聊天服务的插件 | A bot plugin for LLM chat with multi-model integration, extensibility, and various output formats

What it solves

ChatLuna is a flexible chatbot framework designed to integrate various Large Language Models (LLMs) into chat platforms via the Koishi framework. It solves the problem of fragmented model access by providing a unified interface for multiple AI platforms and supporting diverse output formats beyond simple text.

How it works

Built on LangChain and the Koishi API, the project provides a series of adapters for different LLM providers (such as OpenAI, Gemini, Claude, Deepseek, and local models via Ollama or RWKV). It allows users to configure custom personas (presets) using YAML files and supports three distinct operational modes: Chat, Browse (web search), and Agent. It also incorporates content moderation services and long-term memory to maintain context across conversations.

Who it’s for

Users and developers who want to deploy a sophisticated AI chatbot on chat platforms using Koishi, allowing them to switch between different AI models and customize the bot's personality and capabilities.

Highlights

  • Multi-platform Model Support: Integrates with a wide array of global and local LLMs including OpenAI, Azure, Gemini, Claude, and various Chinese models like Qwen and Zhipu.
  • Diverse Output Formats: Supports text, voice (TTS), image rendering, and mixed-media responses.
  • Agentic Capabilities: Includes a dedicated Agent mode and support for the MCP (Model Context Protocol) client.
  • Extensible Architecture: Highly customizable via LangChain and Koishi API, featuring a custom persona system based on YAML presets.
  • Web Search Integration: Provides search capabilities via Google, Bing, DuckDuckGO, and Tavily.