fix(providers): live model discovery for xiaomi-mimo / xiaomi-mimo-token-plan (#15307) - #15308

Problem

xiaomi-mimo and xiaomi-mimo-token-plan are keyed OpenAI-style providers with a real registry baseUrl and a live OpenAI-compatible /v1/models endpoint, but they are not classified by any live-fetch branch of the model-import route (src/app/api/providers/[id]/models/route.ts): they are not openai-compatible-*, not self-hosted, and not in NAMED_OPENAI_STYLE_PROVIDERS. Import therefore never probes the upstream /models and falls through to the hardcoded registry seed (CHAT_OPENAI_COMPAT_MODELS["xiaomi-mimo"] = 2 entries: mimo-v2.5-pro, mimo-v2.5) with the warning API unavailable — using local catalog.

The upstream API itself is healthy — verified from inside an OmniRoute container against the same key:

• GET https://token-plan-sgp.xiaomimimo.com/v1/models → 200 in ~690 ms with 8 models (mimo-v2.5-pro, mimo-v2.5, mimo-v2.6-pro, mimo-v2.6-flash, mimo-v2.5-asr, mimo-v2.5-tts, mimo-v2.5-tts-voiceclone, mimo-v2.5-tts-voicedesign)
• POST /v1/chat/completions → 200 in ~1.1 s

Inference works (DefaultExecutor reads baseUrl from the connection); only model discovery is affected. Syncing consequently imports 2 of 8 models, and newer models (mimo-v2.6-pro, mimo-v2.6-flash) are undiscoverable — they must be added by hand as custom models.

Closes #15307. Same class as #4249 (vercel-ai-gateway), #4202 (zenmux), #3976 (llm7/byteplus) and the provider-model-sweep — each fixed the same way.

Fix

Add "xiaomi-mimo" and "xiaomi-mimo-token-plan" to NAMED_OPENAI_STYLE_PROVIDERS in src/app/api/providers/[id]/models/discovery/providerSets.ts. Live fetch keeps the catalog fresh; the registry seed remains the offline fallback (import never breaks when the upstream is down — covered by the third test).

Testing

New regression test tests/unit/xiaomi-mimo-models-fetch-15307.test.ts (mirrors vercel-gateway-models-fetch-4249.test.ts):

1. both providers are classified by isNamedOpenAIStyleProvider
2. xiaomi-mimo import probes https://api.xiaomimimo.com/v1/models and serves the live catalog (source: "api"), including mimo-v2.6-pro
3. xiaomi-mimo-token-plan import probes https://token-plan-sgp.xiaomimimo.com/v1/models and serves the live catalog
4. when the upstream fetch fails, import falls back to the local catalog (source: "local_catalog") instead of erroring

All 6 tests in the new file + vercel-gateway-models-fetch-4249.test.ts pass locally (node --test, tsx).