feat: add MiniMax as first-class LLM provider

[octo-patch]

Summary

Add MiniMax as a first-class LLM provider for MiroFlow, enabling users to leverage MiniMax M2.7 and M2.7-highspeed models through the OpenAI-compatible API.

Changes

• New provider: MiniMaxClient in miroflow/llm/minimax_client.py – extends LLMClientBase using OpenAI-compatible endpoint at api.minimax.io/v1
• Config template: config/llm/base_minimax.yaml – ready-to-use YAML config for MiniMax M2.7
• Temperature clamping: Automatic clamping to MiniMax valid range (0.0, 1.0]
• Tests: 28 unit tests + 3 integration tests in tests/llm/test_minimax_client.py
• Documentation: Updated README, LLM clients overview, and contribution guide

Supported Models

Model	Description
MiniMax-M2.7	Peak performance, ultimate value (default)
MiniMax-M2.7-highspeed	Same performance, faster and more agile

Both models support 204,800 tokens context window.

Usage

# In your agent config
main_agent:
  llm:
    provider_class: MiniMaxClient
    model_name: MiniMax-M2.7
    api_key: ${oc.env:MINIMAX_API_KEY,???}
    base_url: ${oc.env:MINIMAX_BASE_URL,https://api.minimax.io/v1}

API Documentation

• OpenAI Compatible API

Test Plan

• 28 unit tests covering temperature clamping, client creation, response processing, tool calls, message history
• 3 integration tests with live MiniMax API (M2.7 + M2.7-highspeed)
• All 31 tests passing