- Implement `resolve_user_allowed_model_ids` to enforce role-based model restrictions.
- Add `validate_agent_model_access` to verify model availability and permissions during chat streaming.
- Introduce `AvailableModel` schema to provide safe, filtered model lists to the frontend.
- Add centralized `backendErrors` utility to translate server-side error codes into localized user messages.
- Update `LLMClient` to raise `AuthorizationError` for disabled or missing models.
- Refactor `list_available_models` to return user-specific model subsets.
- Add comprehensive test coverage for model access and authorization logic.
- Update Kubernetes manifests to use secrets for sensitive configuration.