feat: add MiniMax provider support

[ximiximi423]

Summary

• Add MiniMax as a new LLM provider with OpenAI-compatible API
• Support two models: MiniMax-M2.5 (default) and MiniMax-M2.5-highspeed
• Both models support 204,800 tokens context window

Supported Models

Model ID	Type	Description
MiniMax-M2.5	Default	Peak Performance. Ultimate Value. Master the Complex
MiniMax-M2.5-highspeed	Optional	Same performance, faster and more agile

Changes

• Add MiniMaxLLMBackend class (inherits OpenAILLMBackend for OpenAI-compatible API)
• Register MiniMax in both LLM and embedding backend registries in http_client.py
• Add MINIMAX_API_KEY environment variable support with provider defaults in settings.py
• Add __init__.py exports for the new backend
• Add unit tests (tests/llm/test_minimax_provider.py)
• Add integration test (tests/test_minimax.py)

Configuration Example

from memu.app import MemoryService

service = MemoryService(
    llm_profiles={
        "default": {
            "provider": "minimax",
            "base_url": "https://api.minimax.io/v1",
            "api_key": os.getenv("MINIMAX_API_KEY"),
            "chat_model": "MiniMax-M2.5",
        },
    },
)

API Documentation

• OpenAI Compatible API
• Anthropic Compatible API

Test Plan

• Unit tests for backend, settings defaults, payload building, response parsing
• Integration test following the same pattern as OpenRouter tests
• Verified MiniMax backend registration in LLM_BACKENDS and embedding backends
• Verified HTTPLLMClient works with minimax provider