Add MiniMax as a first-class chat LLM provider

[octo-patch]

Summary

Adds MiniMax as a new chat LLM provider alongside OpenAI, Claude, and PaLM.

• MiniMaxChat class extending BaseChatLLM — uses MiniMax's OpenAI-compatible API (https://api.minimax.io/v1), so no new dependencies are required (reuses the existing openai library)
• Supports MiniMax-M2.7 (default) and MiniMax-M2.7-highspeed models with 204K context window
• Temperature clamping to MiniMax's accepted range (0.0, 1.0]
• Per-request api_key and api_base to avoid global state conflicts with OpenAI
• Full sync/async support via the existing call_with_retry / async_call_with_retry utilities

Changes

File	Description
llmflows/llms/minimax_chat.py	New MiniMaxChat class
llmflows/llms/__init__.py	Export MiniMaxChat
tests/llms/test_minimax_chat.py	19 unit tests
tests/llms/test_minimax_chat_integration.py	3 integration tests
README.md	Add MiniMax chat example + mention in features

5 files changed, 567 additions, 1 deletion

Usage

from llmflows.llms import MiniMaxChat, MessageHistory

llm = MiniMaxChat(api_key="<your-minimax-api-key>", model="MiniMax-M2.7")
message_history = MessageHistory()
message_history.add_user_message("Hello!")

response, call_data, model_config = llm.generate(message_history)
print(response)

Test plan

• All 19 unit tests pass (initialization, temperature clamping, generate, format_results, model variants)
• All 3 integration tests pass (multi-turn conversation, max_messages, config completeness)
• All 69 existing tests pass with no regressions
• Manual test with a real MiniMax API key