Skip to content

Add MiniMax M3.1 Flash as the setup default - #95

Open
octo-patch wants to merge 1 commit into
AMAP-ML:mainfrom
octo-patch:octo/20261019-minimax-m31-flash
Open

octo-patch wants to merge 1 commit into
AMAP-ML:mainfrom
octo-patch:octo/20261019-minimax-m31-flash

Conversation

@octo-patch

Copy link
Copy Markdown
Contributor

Reason: The MiniMax setup preset does not expose MiniMax-M3.1-Flash and still defaults new configurations to MiniMax-M3.

Set MiniMax-M3.1-Flash as the default and show MiniMax-M3.1-Flash and MiniMax-M3 before the existing free-form model prompt. Keep saved model IDs, custom model input, and the global/China endpoint selection intact.

Add regression coverage for both regional defaults, the displayed model options, selecting MiniMax-M3, switching to MiniMax, and preserving saved/custom model IDs and endpoints. This is a model-ID preset update; it introduces no pricing, thinking, or service-tier configuration.

Validation:

  • python -m pytest tests/test_setup_wizard.py tests/test_config_store.py --basetemp=.pytest-tmp -q: 11 passed.
  • ruff check . with Ruff 0.12.2: passed.
  • ruff format --check . with Ruff 0.12.2: 78 files already formatted.
  • git diff --check and configured diff secret scan: passed.

Checks used a checkout-local virtual environment with the core package dependencies. No live MiniMax API calls were made.

Official source: https://platform.minimax.io/docs/api-reference/text-chat-openai.md

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant