feat(ai): add Xiaomi MiMo Token Plan support

[xuruiray]

Summary

Add Xiaomi MiMo Token Plan support to the built-in Xiaomi provider and make the model connection check use the same unified LLM wrapper as normal generation.

Related Issues

N/A

Changes

• Added Xiaomi MiMo Token Plan OpenAI-compatible Base URL presets for China, Singapore, and Europe.
• Expanded the built-in Xiaomi MiMo model catalog to include mimo-v2-pro, mimo-v2-omni, and mimo-v2-flash alongside the existing V2.5 models.
• Added thinking metadata for the added MiMo models.
• Updated /api/verify-model to call through callLLM with thinking disabled, so OpenAI-compatible provider request adapters are applied during connection tests.
• Added Xiaomi provider and verify-model route tests.
• Documented MiMo Token Plan setup in .env.example, README.md, and README-zh.md, with locale labels for the new Base URL chips.
• Fixed the OpenAI-compatible fetch wrapper assignment for Next/TypeScript build compatibility.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Refactoring (no functional changes)
• CI/CD or build changes

Verification

Steps to reproduce / test

1. Configure Xiaomi Token Plan credentials with MIMO_API_KEY=tp-..., MIMO_BASE_URL=https://token-plan-cn.xiaomimimo.com/v1, and DEFAULT_MODEL=xiaomi:mimo-v2.5-pro.
2. Use the settings model connection test or any server-side LLM call path using xiaomi:mimo-v2.5-pro.
3. Confirm the model returns text successfully and the Token Plan endpoint is used.

What I personally verified

• pnpm vitest run tests/ai/xiaomi-provider.test.ts tests/api/verify-model.test.ts tests/ai/openai-provider.test.ts tests/server/provider-config.test.ts tests/ai/thinking-config.test.ts
  • Result: 5 test files passed, 62 tests passed.
• pnpm test
  • Result: 47 test files passed, 338 tests passed.
• pnpm check:i18n-keys
  • Result: passed for 6 locale files.
• pnpm check
  • Result: all matched files use Prettier style.
• pnpm lint
  • Result: 0 errors, 7 existing warnings in unrelated files.
• pnpm build
  • Result: compiled successfully; Next emitted existing workspace-root and middleware convention warnings.
• Real Xiaomi MiMo Token Plan smoke via getModel + callLLM against https://token-plan-cn.xiaomimimo.com/v1, using mimo-v2.5-pro and thinking disabled.
  • Sanitized result: text=MIMO_OK, inputTokens=263, outputTokens=4, reasoningTokens=0.

Evidence

• CI passes (pnpm check && pnpm lint && npx tsc --noEmit)
• Manually tested locally
• Screenshots / recordings attached (if UI changes)

Checklist

• My code follows the project's coding style
• I have performed a self-review of my code
• I have added/updated documentation as needed
• My changes do not introduce new warnings

[wyuc]

LGTM. Verified the model-metadata wiring and the verify-model refactor; also checked the settings UI locally (Token Plan chips + 5 models render fine, no component change).

No Token Plan key on my side, so no e2e against the real endpoints — relying on the author's smoke test + green CI.

Approving.