Add CometAPI LLM Provider Support

[tensornull]

Hi MetaGPT team,

We're CometAPI, and we offer a robust API that provides access to 500+ large language models (LLMs), as well as multimodal generation capabilities (text, images, video, audio, etc.). Our platform makes it easy to integrate these advanced features into your project, empowering developers with powerful AI tools.

MetaGPT already supports 27 LLM types (OpenAI, Anthropic, Azure, Gemini, Ollama, Zhipuai, Qianfan, DashScope, DeepSeek, and many more) through 14+ provider implementations in metagpt/provider/. Adding CometAPI would give MetaGPT users access to 500+ models through a single, unified API, making it easier to:

• Assign specialized models to different agent roles (ProductManager, Architect, Engineer, QA) without managing multiple API keys
• Access models from OpenAI, Anthropic, Google, DeepSeek, Alibaba, and others through one endpoint
• Test and compare different models for multi-agent workflows
• Benefit from competitive pricing with unified billing across all models
• Leverage multimodal capabilities (image/video/audio generation) for enhanced agent outputs

Since CometAPI uses OpenAI-compatible endpoints, integration is straightforward. Based on our code analysis of metagpt/provider/openai_api.py, MetaGPT's OpenAILLM class already supports custom base_url, so users can start using CometAPI immediately with existing configuration:

llm:
  api_type: openai
  api_key: YOUR_COMETAPI_KEY
  base_url: https://api.cometapi.com/v1/
  model: gpt-5.1  # or claude-4-5-sonnet, gemini-3.0-pro-preview, etc.

For better user experience, we'd recommend adding COMETAPI = "cometapi" to the LLMType enum in metagpt/configs/llm_config.py and registering it in the @register_provider decorator.

CometAPI Resources

• Website
• Documentation
• Get an API Key
• Pricing

Implementation Offer
We'd be glad to help integrate CometAPI into MetaGPT. If you're interested, we can submit a pull request that aligns with your project's coding standards and guidelines.

---