dynamic LLM provider abstraction proxy

infrastructureexternal call

UnifiedChatCompletionRequest -> ProviderSpecificResponse

Map a generic model alias in a chat completion request to a specific configured provider backend and execute the call.

Problem it solves

Hardcoding provider SDKs binds application code to specific LLM APIs and requires code redeployment to change models.

Consumes

UnifiedChatCompletionRequest

Emits

ProviderSpecificResponse

Distilled from 1 source

The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.