fix(openai.py): support completion, streaming, async_streaming

litellm/llms/openai.py

@@ -803,6 +803,7 @@ class OpenAIChatCompletion(BaseLLM):
         client=None,
         organization: Optional[str] = None,
         custom_llm_provider: Optional[str] = None,
+        drop_params: Optional[bool] = None,
     ):
         super().completion()
         exception_mapping_worked = False
@@ -858,6 +859,7 @@ class OpenAIChatCompletion(BaseLLM):
                                 client=client,
                                 max_retries=max_retries,
                                 organization=organization,
+                                drop_params=drop_params,
                             )
                         else:
                             return self.acompletion(
@@ -871,6 +873,7 @@ class OpenAIChatCompletion(BaseLLM):
                                 client=client,
                                 max_retries=max_retries,
                                 organization=organization,
+                                drop_params=drop_params,
                             )
                     elif optional_params.get("stream", False):
                         return self.streaming(
@@ -925,6 +928,33 @@ class OpenAIChatCompletion(BaseLLM):
                             response_object=stringified_response,
                             model_response_object=model_response,
                         )
+                except openai.UnprocessableEntityError as e:
+                    ## check if body contains unprocessable params - related issue https://github.com/BerriAI/litellm/issues/4800
+                    if litellm.drop_params is True or drop_params is True:
+                        if e.body is not None and e.body.get("detail"):  # type: ignore
+                            detail = e.body.get("detail")  # type: ignore
+                            invalid_params: List[str] = []
+                            if (
+                                isinstance(detail, List)
+                                and len(detail) > 0
+                                and isinstance(detail[0], dict)
+                            ):
+                                for error_dict in detail:
+                                    if (
+                                        error_dict.get("loc")
+                                        and isinstance(error_dict.get("loc"), list)
+                                        and len(error_dict.get("loc")) == 2
+                                    ):
+                                        invalid_params.append(error_dict["loc"][1])
+
+                        new_data = {}
+                        for k, v in optional_params.items():
+                            if k not in invalid_params:
+                                new_data[k] = v
+                        optional_params = new_data
+                    else:
+                        raise e
+                    # e.message
                 except Exception as e:
                     if print_verbose is not None:
                         print_verbose(f"openai.py: Received openai error - {str(e)}")
@@ -982,6 +1012,7 @@ class OpenAIChatCompletion(BaseLLM):
         client=None,
         max_retries=None,
         headers=None,
+        drop_params: Optional[bool] = None,
     ):
         response = None
         for _ in range(
@@ -1030,7 +1061,7 @@ class OpenAIChatCompletion(BaseLLM):
                 )
             except openai.UnprocessableEntityError as e:
                 ## check if body contains unprocessable params - related issue https://github.com/BerriAI/litellm/issues/4800
-                if litellm.drop_params is True:
+                if litellm.drop_params is True or drop_params is True:
                     if e.body is not None and e.body.get("detail"):  # type: ignore
                         detail = e.body.get("detail")  # type: ignore
                         invalid_params: List[str] = []
@@ -1113,57 +1144,87 @@ class OpenAIChatCompletion(BaseLLM):
         client=None,
         max_retries=None,
         headers=None,
+        drop_params: Optional[bool] = None,
     ):
         response = None
-        try:
+        for _ in range(2):
-            openai_aclient = self._get_openai_client(
+            try:
-                is_async=True,
+                openai_aclient = self._get_openai_client(
-                api_key=api_key,
+                    is_async=True,
-                api_base=api_base,
+                    api_key=api_key,
-                timeout=timeout,
+                    api_base=api_base,
-                max_retries=max_retries,
+                    timeout=timeout,
-                organization=organization,
+                    max_retries=max_retries,
-                client=client,
+                    organization=organization,
-            )
+                    client=client,
-            ## LOGGING
-            logging_obj.pre_call(
-                input=data["messages"],
-                api_key=api_key,
-                additional_args={
-                    "headers": headers,
-                    "api_base": api_base,
-                    "acompletion": True,
-                    "complete_input_dict": data,
-                },
-            )
-
-            headers, response = await self.make_openai_chat_completion_request(
-                openai_aclient=openai_aclient, data=data, timeout=timeout
-            )
-            logging_obj.model_call_details["response_headers"] = headers
-            streamwrapper = CustomStreamWrapper(
-                completion_stream=response,
-                model=model,
-                custom_llm_provider="openai",
-                logging_obj=logging_obj,
-                stream_options=data.get("stream_options", None),
-            )
-            return streamwrapper
-        except (
-            Exception
-        ) as e:  # need to exception handle here. async exceptions don't get caught in sync functions.
-            if response is not None and hasattr(response, "text"):
-                raise OpenAIError(
-                    status_code=500,
-                    message=f"{str(e)}\n\nOriginal Response: {response.text}",
                 )
-            else:
+                ## LOGGING
-                if type(e).__name__ == "ReadTimeout":
+                logging_obj.pre_call(
-                    raise OpenAIError(status_code=408, message=f"{type(e).__name__}")
+                    input=data["messages"],
-                elif hasattr(e, "status_code"):
+                    api_key=api_key,
-                    raise OpenAIError(status_code=e.status_code, message=str(e))
+                    additional_args={
+                        "headers": headers,
+                        "api_base": api_base,
+                        "acompletion": True,
+                        "complete_input_dict": data,
+                    },
+                )
+
+                headers, response = await self.make_openai_chat_completion_request(
+                    openai_aclient=openai_aclient, data=data, timeout=timeout
+                )
+                logging_obj.model_call_details["response_headers"] = headers
+                streamwrapper = CustomStreamWrapper(
+                    completion_stream=response,
+                    model=model,
+                    custom_llm_provider="openai",
+                    logging_obj=logging_obj,
+                    stream_options=data.get("stream_options", None),
+                )
+                return streamwrapper
+            except openai.UnprocessableEntityError as e:
+                ## check if body contains unprocessable params - related issue https://github.com/BerriAI/litellm/issues/4800
+                if litellm.drop_params is True or drop_params is True:
+                    if e.body is not None and e.body.get("detail"):  # type: ignore
+                        detail = e.body.get("detail")  # type: ignore
+                        invalid_params: List[str] = []
+                        if (
+                            isinstance(detail, List)
+                            and len(detail) > 0
+                            and isinstance(detail[0], dict)
+                        ):
+                            for error_dict in detail:
+                                if (
+                                    error_dict.get("loc")
+                                    and isinstance(error_dict.get("loc"), list)
+                                    and len(error_dict.get("loc")) == 2
+                                ):
+                                    invalid_params.append(error_dict["loc"][1])
+
+                    new_data = {}
+                    for k, v in data.items():
+                        if k not in invalid_params:
+                            new_data[k] = v
+                    data = new_data
                 else:
-                    raise OpenAIError(status_code=500, message=f"{str(e)}")
+                    raise e
+            except (
+                Exception
+            ) as e:  # need to exception handle here. async exceptions don't get caught in sync functions.
+                if response is not None and hasattr(response, "text"):
+                    raise OpenAIError(
+                        status_code=500,
+                        message=f"{str(e)}\n\nOriginal Response: {response.text}",
+                    )
+                else:
+                    if type(e).__name__ == "ReadTimeout":
+                        raise OpenAIError(
+                            status_code=408, message=f"{type(e).__name__}"
+                        )
+                    elif hasattr(e, "status_code"):
+                        raise OpenAIError(status_code=e.status_code, message=str(e))
+                    else:
+                        raise OpenAIError(status_code=500, message=f"{str(e)}")
 
     # Embedding
     async def make_openai_embedding_request(

litellm/main.py

@@ -1176,6 +1176,7 @@ def completion(
                     client=client,  # pass AsyncOpenAI, OpenAI client
                     organization=organization,
                     custom_llm_provider=custom_llm_provider,
+                    drop_params=non_default_params.get("drop_params"),
                 )
             except Exception as e:
                 ## LOGGING - log the original exception returned

litellm/tests/test_completion.py

@@ -142,6 +142,36 @@ def test_completion_azure_ai_command_r():
         pytest.fail(f"Error occurred: {e}")
 
 
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_completion_azure_ai_mistral_invalid_params(sync_mode):
+    try:
+        import os
+
+        litellm.set_verbose = True
+
+        os.environ["AZURE_AI_API_BASE"] = os.getenv("AZURE_MISTRAL_API_BASE", "")
+        os.environ["AZURE_AI_API_KEY"] = os.getenv("AZURE_MISTRAL_API_KEY", "")
+
+        data = {
+            "model": "azure_ai/mistral",
+            "messages": [{"role": "user", "content": "What is the meaning of life?"}],
+            "frequency_penalty": 0.1,
+            "presence_penalty": 0.1,
+            "drop_params": True,
+        }
+        if sync_mode:
+            response: litellm.ModelResponse = completion(**data)  # type: ignore
+        else:
+            response: litellm.ModelResponse = await litellm.acompletion(**data)  # type: ignore
+
+        assert "azure_ai" in response.model
+    except litellm.Timeout as e:
+        pass
+    except Exception as e:
+        pytest.fail(f"Error occurred: {e}")
+
+
 def test_completion_azure_command_r():
     try:
         litellm.set_verbose = True

litellm/tests/test_streaming.py

@@ -2881,6 +2881,40 @@ def test_azure_streaming_and_function_calling():
         raise e
 
 
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_completion_azure_ai_mistral_invalid_params(sync_mode):
+    try:
+        import os
+
+        litellm.set_verbose = True
+
+        os.environ["AZURE_AI_API_BASE"] = os.getenv("AZURE_MISTRAL_API_BASE", "")
+        os.environ["AZURE_AI_API_KEY"] = os.getenv("AZURE_MISTRAL_API_KEY", "")
+
+        data = {
+            "model": "azure_ai/mistral",
+            "messages": [{"role": "user", "content": "What is the meaning of life?"}],
+            "frequency_penalty": 0.1,
+            "presence_penalty": 0.1,
+            "drop_params": True,
+            "stream": True,
+        }
+        if sync_mode:
+            response: litellm.ModelResponse = completion(**data)  # type: ignore
+            for chunk in response:
+                print(chunk)
+        else:
+            response: litellm.ModelResponse = await litellm.acompletion(**data)  # type: ignore
+
+            async for chunk in response:
+                print(chunk)
+    except litellm.Timeout as e:
+        pass
+    except Exception as e:
+        pytest.fail(f"Error occurred: {e}")
+
+
 @pytest.mark.asyncio
 async def test_azure_astreaming_and_function_calling():
     import uuid