fix llm request utils

litellm/litellm_core_utils/llm_request_utils.py

@@ -3,89 +3,66 @@ from typing import Dict, Optional
 import litellm
 
 
-class LitellmCoreRequestUtils:
+def _ensure_extra_body_is_safe(extra_body: Optional[Dict]) -> Optional[Dict]:
+    """
+    Ensure that the extra_body sent in the request is safe,  otherwise users will see this error
 
-    @staticmethod
-    def _ensure_extra_body_is_safe(extra_body: Optional[Dict]) -> Optional[Dict]:
-        """
-        Ensure that the extra_body sent in the request is safe,  otherwise users will see this error
-
-        "Object of type TextPromptClient is not JSON serializable
+    "Object of type TextPromptClient is not JSON serializable
 
 
-        Relevant Issue: https://github.com/BerriAI/litellm/issues/4140
-        """
-        if extra_body is None:
-            return None
-
-        if not isinstance(extra_body, dict):
-            return extra_body
-
-        if "metadata" in extra_body and isinstance(extra_body["metadata"], dict):
-            if "prompt" in extra_body["metadata"]:
-                _prompt = extra_body["metadata"].get("prompt")
-
-                # users can send Langfuse TextPromptClient objects, so we need to convert them to dicts
-                # Langfuse TextPromptClients have .__dict__ attribute
-                if _prompt is not None and hasattr(_prompt, "__dict__"):
-                    extra_body["metadata"]["prompt"] = _prompt.__dict__
+    Relevant Issue: https://github.com/BerriAI/litellm/issues/4140
+    """
+    if extra_body is None:
+        return None
 
+    if not isinstance(extra_body, dict):
         return extra_body
 
-    @staticmethod
-    def pick_cheapest_chat_models_from_llm_provider(custom_llm_provider: str, n=1):
-        """
-        Pick the n cheapest chat models from the LLM provider.
+    if "metadata" in extra_body and isinstance(extra_body["metadata"], dict):
+        if "prompt" in extra_body["metadata"]:
+            _prompt = extra_body["metadata"].get("prompt")
 
-        Args:
-            custom_llm_provider (str): The name of the LLM provider.
-            n (int): The number of cheapest models to return.
+            # users can send Langfuse TextPromptClient objects, so we need to convert them to dicts
+            # Langfuse TextPromptClients have .__dict__ attribute
+            if _prompt is not None and hasattr(_prompt, "__dict__"):
+                extra_body["metadata"]["prompt"] = _prompt.__dict__
 
-        Returns:
-            list[str]: A list of the n cheapest chat models.
-        """
-        if custom_llm_provider not in litellm.models_by_provider:
-            return []
+    return extra_body
 
-        known_models = litellm.models_by_provider.get(custom_llm_provider, [])
-        model_costs = []
 
-        for model in known_models:
-            try:
-                model_info = litellm.get_model_info(
-                    model=model, custom_llm_provider=custom_llm_provider
-                )
-            except Exception:
-                continue
-            if model_info.get("mode") != "chat":
-                continue
-            _cost = model_info.get("input_cost_per_token", 0) + model_info.get(
-                "output_cost_per_token", 0
+def pick_cheapest_chat_models_from_llm_provider(custom_llm_provider: str, n=1):
+    """
+    Pick the n cheapest chat models from the LLM provider.
+
+    Args:
+        custom_llm_provider (str): The name of the LLM provider.
+        n (int): The number of cheapest models to return.
+
+    Returns:
+        list[str]: A list of the n cheapest chat models.
+    """
+    if custom_llm_provider not in litellm.models_by_provider:
+        return []
+
+    known_models = litellm.models_by_provider.get(custom_llm_provider, [])
+    model_costs = []
+
+    for model in known_models:
+        try:
+            model_info = litellm.get_model_info(
+                model=model, custom_llm_provider=custom_llm_provider
             )
-            model_costs.append((model, _cost))
+        except Exception:
+            continue
+        if model_info.get("mode") != "chat":
+            continue
+        _cost = model_info.get("input_cost_per_token", 0) + model_info.get(
+            "output_cost_per_token", 0
+        )
+        model_costs.append((model, _cost))
 
-        # Sort by cost (ascending)
-        model_costs.sort(key=lambda x: x[1])
+    # Sort by cost (ascending)
+    model_costs.sort(key=lambda x: x[1])
 
-        # Return the top n cheapest models
-        return [model for model, _ in model_costs[:n]]
-
-    @staticmethod
-    def select_model_for_request_transformation(
-        model: str,
-        base_model: Optional[str] = None,
-        litellm_params: Optional[Dict] = None,
-    ) -> str:
-        """
-        If `base_model` is passed in by user, use it for the request transformation
-
-        Else, use the model passed in the request
-        """
-        if base_model is not None:
-            return base_model
-        elif (
-            litellm_params is not None and litellm_params.get("base_model") is not None
-        ):
-            return litellm_params["base_model"]
-        else:
-            return model
+    # Return the top n cheapest models
+    return [model for model, _ in model_costs[:n]]