This project provides a comprehensive backend implementation for FeiShu (Lark) chatbots, enabling users to interact with and control advanced AI agents directly within their chat interface. It aims to offer a natural conversational experience by leveraging FeiShu's unique features like streaming output and integrated user management, making it an ideal solution for users seeking to harness multi-agent AI capabilities through a familiar chat environment.

How It Works

The system acts as a unified AI server, integrating numerous large language models (LLMs) and providing a single API for model interaction. It supports complex agent functionalities, including web scraping, function calling for tools like weather APIs and search engines, image generation with models like Midjourney, and code execution for calculations. Advanced agents can orchestrate multiple LLMs and tools, with specific models like GPT-4o and Claude 3.7 recommended for more complex tasks and features like virtual browser automation.

Quick Start & Requirements

• Installation: Docker is mandatory. A server with a publicly accessible domain is required.
• Prerequisites:
  • Docker and Docker Compose.
  • A created FeiShu (Lark) bot with appropriate permissions and subscribed events (im.message.receive_v1, application.bot.menu_v6, card.action.trigger).
  • API keys for desired LLM providers (e.g., OpenAI, MiniMax, Gemini).
  • A MySQL database and Redis instance.
• Setup: Clone the repository, configure config.json with FeiShu bot credentials and LLM API keys, set up MySQL and Redis (Docker examples provided), build the WebSocket service (sh buildwebsocket.sh), and run the main Docker container (docker run -p 8080:8080 ai_proxy). FeiShu bot event callbacks should be configured for long polling.

Highlighted Details

• Supports rendering Mermaid diagrams, Markmap mind maps, LaTeX formulas, and HTML code directly within FeiShu cards.
• Enables interactive charts from data returned by LLMs.
• Offers advanced agents for automated book summarization, web automation via a virtual browser (requires Claude 3.7), and complex task orchestration.
• Provides a unified Web API for integration with web, mini-programs, and native apps, including text-to-speech and unified WebSocket interfaces for real-time voice and translation.

Maintenance & Community

The project is actively developed by unfish. Community interaction and discussion are encouraged via a FeiShu group.

Licensing & Compatibility

The repository does not explicitly state a license in the README. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The project is presented as a code-sharing initiative, requiring users to manage their own LLM accounts and API keys. Some advanced features, like virtual browser automation, have specific model dependencies (Claude 3.7). The effectiveness of complex agents is highly dependent on the underlying LLM's capabilities.