feat: add MiniMax chat model support

[octo-patch]

Summary

Adds MiniMaxChatModel as a dedicated model class for MiniMax's LLM API, enabling first-class support for MiniMax models in AgentScope.

Changes

• src/agentscope/model/_minimax_model.py: New model class extending OpenAIChatModel with MiniMax-specific defaults (base URL, API key env var, model names)
• src/agentscope/model/__init__.py: Added MiniMaxChatModel to module exports

Usage

from agentscope.model import MiniMaxChatModel

model = MiniMaxChatModel(
    model_name="MiniMax-M2.5",
    api_key="your-api-key",  # or set MINIMAX_API_KEY env var
)

Key Details

• Extends OpenAIChatModel since MiniMax API is OpenAI-compatible
• Default base URL: https://api.minimax.io/v1
• Default model: MiniMax-M2.5 (204K context window)
• Also supports MiniMax-M2.5-highspeed for speed-optimized inference
• Reads API key from MINIMAX_API_KEY environment variable
• Temperature note: MiniMax does not accept exactly 0; use small positive values

Test Plan

• Import MiniMaxChatModel from agentscope.model
• Verify default base URL is set to MiniMax endpoint
• Test chat completion with MiniMax-M2.5
• Test streaming mode
• Verify MINIMAX_API_KEY env var is read when api_key not provided

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[qbc2016]

Thanks for this PR!

There's already a similar PR (#1302) that implements this feature. Since theirs was opened first, we'll proceed with reviewing that one.

If you have ideas for improvements or a different implementation approach, feel free to share your thoughts on PR #1302 or the related discussion. We value collaboration and would love to hear your perspective!

Thanks again for contributing!

To keep the discussion and review in one place, I’m going to close this one.