Add MiniMax as first-class chat model provider

[octo-patch]

Summary

• Add MiniMaxChatTarget as a new prompt target for MiniMax AI chat models via OpenAI-compatible API
• Support MiniMax-M2.7, M2.7-highspeed, M2.5, and M2.5-highspeed models (up to 1M token context)
• Includes temperature clamping [0, 1], <think> tag stripping for reasoning models, JSON mode, and token usage tracking
• 40 unit tests + 3 integration tests
• Environment variable configuration in .env_example and documentation in prompt targets overview

Details

New Files

• pyrit/prompt_target/minimax/__init__.py - Package init
• pyrit/prompt_target/minimax/minimax_chat_target.py - Main implementation (~250 lines)
• tests/unit/target/test_minimax_chat_target.py - 40 unit tests
• tests/integration/targets/test_minimax_chat_target_integration.py - 3 integration tests

Modified Files

• pyrit/prompt_target/__init__.py - Register MiniMaxChatTarget
• .env_example - Add MINIMAX_API_KEY, MINIMAX_CHAT_ENDPOINT, MINIMAX_CHAT_MODEL
• doc/code/targets/0_prompt_targets.md - Add MiniMax to targets table

Implementation Approach

MiniMaxChatTarget extends PromptChatTarget directly (rather than OpenAITarget) to provide a clean, focused implementation:

• Uses AsyncOpenAI client pointing at https://api.minimax.io/v1 (OpenAI-compatible)
• Temperature clamping: MiniMax accepts [0, 1] range vs OpenAI's [0, 2]; values > 1.0 are clamped with a warning
• Think tag stripping: MiniMax M2.5+ models may include <think>...</think> reasoning tags; these are automatically stripped
• JSON mode: Supports json_object response format (falls back from json_schema since MiniMax does not support structured output schemas)
• Environment variables: MINIMAX_API_KEY (required), MINIMAX_CHAT_ENDPOINT (optional), MINIMAX_CHAT_MODEL (optional)

Usage Example

from pyrit.prompt_target import MiniMaxChatTarget

target = MiniMaxChatTarget(
    api_key="your-minimax-api-key",
    model_name="MiniMax-M2.7",
    temperature=0.7,
)

Test Plan

• 40 unit tests covering initialization, validation, request building, message construction, error handling, and token usage
• 3 integration tests covering basic completion, multi-turn conversation, and JSON mode (require MINIMAX_API_KEY)
• All unit tests pass locally