Change model, endpoint, and credentials as one routing unit

When switching an OpenAI-compatible LLM provider, treat the model name, base URL, credential source, and provider-selection flag as one atomic routing unit. Changing only the model can leave a native OpenAI client pointed at OpenAI while requesting a private model, producing immediate 404 model-not-found failures even though the private inference service is healthy. Verify the effective runtime tuple in startup logs and one real application request before expanding traffic; control-plane health probes alone cannot catch a bypassed route.

Applies when: moving a deployed application between native OpenAI and an OpenAI-compatible proxy. Keywords: OpenAI compatible proxy model base URL credentials Railway routing deployment