Introduce Kaapi-Abstracted LLM Parameters to Support Multiple Providers

[avirajsingh7]

Currently, the LLM call endpoint directly accepts and forwards parameters as-is to OpenAI.

We should introduce Kaapi-abstracted LLM parameters that provide a unified interface across multiple LLM providers. Kaapi will be responsible for mapping these abstracted parameters to provider-specific parameters in the backend.

This will significantly simplify the user experience when switching between models and providers.