Skip to content

feat: upgrade MiniMax provider to M2.7 with correct API base URL#85

Open
octo-patch wants to merge 1 commit into
OpenCoworkAI:mainfrom
octo-patch:feature/upgrade-minimax-m27
Open

feat: upgrade MiniMax provider to M2.7 with correct API base URL#85
octo-patch wants to merge 1 commit into
OpenCoworkAI:mainfrom
octo-patch:feature/upgrade-minimax-m27

Conversation

@octo-patch
Copy link
Copy Markdown

Summary

Upgrade MiniMax provider configuration from the outdated M2.5 + api.minimax.chat to the current M2.7 + api.minimax.io.

Changes

  • Base URL: Updated from https://api.minimax.chat/v1 to https://api.minimax.io/v1 (the current official API domain)
  • Models: Replaced MiniMax-M2.5 with MiniMax-M2.7 and added MiniMax-M2.7-highspeed (204K context window)
  • URL matcher: Added api.minimax.io to the hosts array so auto-detection works for both old and new domains
  • README (EN + ZH): Fixed the MiniMax row — corrected base URL, model name, and platform link
  • Docs: Updated docs/common-provider-setups.md with current info
  • Tests: Added 3 unit tests for MiniMax URL detection (new domain, legacy domain, subdomain)

Files changed (7)

File Change
src/shared/api-provider-guidance.ts Updated base URL, model, and hosts matcher
src/shared/api-model-presets.ts M2.5 → M2.7 + M2.7-highspeed in custom presets
docs/common-provider-setups.md Updated MiniMax row
readme.md Fixed MiniMax entry in Quick Start table
README_zh.md Fixed MiniMax entry in Quick Start table
tests/provider-guidance.test.ts Added 3 MiniMax detection tests
tests/api-config-state.test.ts Updated preset assertion for new model IDs

Test plan

  • npx vitest run tests/provider-guidance.test.ts — all 10 tests pass (including 3 new MiniMax tests)
  • npx vitest run tests/api-config-state.test.ts — all 15 tests pass
  • Full test suite: all pre-existing failures are unrelated (timeout issues in store-encryption, remote-manager, etc.)
  • Manual: verify detectCommonProviderSetup('https://api.minimax.io/v1') returns the MiniMax setup in the UI

Context: MiniMax migrated their API from api.minimax.chat to api.minimax.io, and released M2.7 as their latest flagship model (204K context). This PR ensures Open Cowork users get the correct setup guidance.

- Update base URL from api.minimax.chat/v1 to api.minimax.io/v1
- Add M2.7 and M2.7-highspeed models to custom provider presets
- Add api.minimax.io to URL matcher hosts for auto-detection
- Fix outdated MiniMax entries in both English and Chinese READMEs
- Add unit tests for MiniMax URL detection (both old and new domains)
@hqhq1025 hqhq1025 closed this Apr 13, 2026
@hqhq1025 hqhq1025 reopened this Apr 13, 2026
@hqhq1025
Copy link
Copy Markdown
Collaborator

Hi @octo-patch, thanks for this PR! I've reviewed the changes manually (Codex auto-review couldn't run due to merge conflicts).

Overall: Looks good — MiniMax M2.7 upgrade with proper backward compat for the old api.minimax.chat domain. Tests are solid.

One minor note: The MiniMax link in both READMEs changed from the specific Coding Plan page to just the platform homepage. If the Coding Plan page still exists, it might be worth keeping the direct link for users.

Action needed: This PR has merge conflicts with main — could you rebase? Once conflicts are resolved, Codex AI review will run automatically and we can merge.


Manual review by @hqhq1025

@hqhq1025 hqhq1025 added bot-rerun Temporary label for rerunning bot automation and removed bot-rerun Temporary label for rerunning bot automation labels Apr 27, 2026
Copy link
Copy Markdown

@github-actions github-actions Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Findings

  • [Major] MiniMax model IDs were added to presets.custom, but the recommended MiniMax flow is Custom + OpenAI, and that profile reads its options from presets.openai instead. The new entries therefore do not populate the model picker for the flow this PR updates, and the added test only validates the unused preset bucket. Evidence: src/shared/api-model-presets.ts:111, related context src/renderer/hooks/useApiConfigState.ts:178 and src/renderer/hooks/useApiConfigState.ts:547.

Summary

Review mode: initial

1 major issue found. The MiniMax provider guidance and docs now point users to Custom + OpenAI, but this preset change still leaves MiniMax-M2.7 and MiniMax-M2.7-highspeed out of that picker, so users fall back to manual entry. Residual risk: there is still no end-to-end test that applies the detected MiniMax setup and verifies the chosen model state.

Testing

Not run (automation)

Open Cowork Bot

{ id: 'kimi-k2-thinking', name: 'kimi-k2-thinking' },
{ id: 'glm-5', name: 'glm-5' },
{ id: 'MiniMax-M2.5', name: 'MiniMax-M2.5' },
{ id: 'MiniMax-M2.7', name: 'MiniMax-M2.7' },
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

[MAJOR] These model IDs are being added to presets.custom, but the MiniMax setup you detect recommends Custom + OpenAI, and custom:openai resolves its model options from presets.openai (src/renderer/hooks/useApiConfigState.ts:178, src/renderer/hooks/useApiConfigState.ts:547). That means applying the MiniMax setup still drops the user into manual-model mode instead of surfacing these IDs in the picker.

Suggested fix:

openai: {
  // ...
  models: [
    // existing OpenAI-compatible models...
    { id: 'MiniMax-M2.7', name: 'MiniMax-M2.7' },
    { id: 'MiniMax-M2.7-highspeed', name: 'MiniMax-M2.7-highspeed' },
  ],
}

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.

2 participants