feat(models): add MiniMax model support

[ximiximi423]

Summary

• Add support for MiniMax-M2.5 and MiniMax-M2.5-highspeed chat models via OpenAI-compatible API
• Add support for MiniMax TTS models (speech-2.8-hd, speech-2.8-turbo, etc.) via T2A v2 HTTP API
• 17 comprehensive unit tests (9 chat model + 8 TTS)

Chat Model Details

MiniMax-M2.5 Model Features

• OpenAI-compatible API: Uses the openai SDK with MiniMax base URL
• Interleaved Thinking: Parses <think>...</think> tags into ThinkingBlock objects
• Tool Calling: Full support for function calling with tools and tool_choice
• Streaming: Both streaming and non-streaming modes
• Structured Output: Pydantic BaseModel support

Usage Example

from agentscope.model import MiniMaxChatModel
model = MiniMaxChatModel(model_name="MiniMax-M2.5", api_key="your-key")

TTS Model Details

MiniMax TTS Features

• T2A v2 HTTP API: Direct HTTP calls via httpx
• Multiple Models: speech-2.8-hd (default), speech-2.8-turbo, etc.
• Streaming: Newline-delimited JSON streaming
• Voice Settings: Configurable voice_id, speed, volume, pitch, emotion
• Audio Formats: mp3, pcm, flac, wav

Usage Example

from agentscope.tts import MiniMaxTTSModel
tts = MiniMaxTTSModel(api_key="your-key", model_name="speech-2.8-hd")

Test plan

• All 17 new unit tests pass
• All 373 total tests pass (0 failures)

[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]

Thank you for the contribution!

The MiniMax official documentation recommends using the Anthropic-compatible API, so I wonder if we could just use AnthropicChatModel and AnthropicFormatter instead?