Litellm dev 12 17 2024 p2 (#7277)

* fix(openai/transcription/handler.py): call 'log_pre_api_call' on async calls

* fix(openai/transcriptions/handler.py): call 'logging.pre_call' on sync whisper calls as well

* fix(proxy_cli.py): remove default proxy_cli timeout param

gets passed in as a dynamic request timeout and overrides config values

* fix(langfuse.py): pass litellm httpx client - contains ssl certs (#7052)

Fixes https://github.com/BerriAI/litellm/issues/7046

litellm/integrations/langfuse/langfuse.py

@@ -13,6 +13,7 @@ from pydantic import BaseModel
 import litellm
 from litellm._logging import verbose_logger
 from litellm.litellm_core_utils.redact_messages import redact_user_api_key_info
+from litellm.llms.custom_httpx.http_handler import _get_httpx_client
 from litellm.secret_managers.main import str_to_bool
 from litellm.types.integrations.langfuse import *
 from litellm.types.utils import StandardLoggingPayload
@@ -56,6 +57,8 @@ class LangFuseLogger:
         self.langfuse_flush_interval = (
             os.getenv("LANGFUSE_FLUSH_INTERVAL") or flush_interval
         )
+        http_client = _get_httpx_client()
+        self.langfuse_client = http_client.client
 
         parameters = {
             "public_key": self.public_key,
@@ -64,6 +67,7 @@ class LangFuseLogger:
             "release": self.langfuse_release,
             "debug": self.langfuse_debug,
             "flush_interval": self.langfuse_flush_interval,  # flush interval in seconds
+            "httpx_client": self.langfuse_client,
         }
 
         if Version(langfuse.version.__version__) >= Version("2.6.0"):

litellm/llms/openai/transcriptions/handler.py

@@ -9,6 +9,7 @@ from litellm.litellm_core_utils.audio_utils.utils import get_audio_file_name
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.types.utils import FileTypes
 from litellm.utils import TranscriptionResponse, convert_to_model_response_object
+
 from ..openai import OpenAIChatCompletion
 
 
@@ -103,6 +104,17 @@ class OpenAIAudioTranscription(OpenAIChatCompletion):
             timeout=timeout,
             max_retries=max_retries,
         )
+
+        ## LOGGING
+        logging_obj.pre_call(
+            input=None,
+            api_key=openai_client.api_key,
+            additional_args={
+                "api_base": openai_client._base_url._uri_reference,
+                "atranscription": True,
+                "complete_input_dict": data,
+            },
+        )
         _, response = self.make_sync_openai_audio_transcriptions_request(
             openai_client=openai_client,
             data=data,
@@ -147,6 +159,16 @@ class OpenAIAudioTranscription(OpenAIChatCompletion):
                 client=client,
             )
 
+            ## LOGGING
+            logging_obj.pre_call(
+                input=None,
+                api_key=openai_aclient.api_key,
+                additional_args={
+                    "api_base": openai_aclient._base_url._uri_reference,
+                    "atranscription": True,
+                    "complete_input_dict": data,
+                },
+            )
             headers, response = await self.make_openai_audio_transcriptions_request(
                 openai_aclient=openai_aclient,
                 data=data,

litellm/main.py

@@ -4632,6 +4632,11 @@ def transcription(
 
     Allows router to load balance between them
     """
+    litellm_call_id = kwargs.get("litellm_call_id", None)
+    proxy_server_request = kwargs.get("proxy_server_request", None)
+    model_info = kwargs.get("model_info", None)
+    metadata = kwargs.get("metadata", None)
+    atranscription = kwargs.get("atranscription", False)
     atranscription = kwargs.get("atranscription", False)
     litellm_logging_obj: LiteLLMLoggingObj = kwargs.get("litellm_logging_obj")  # type: ignore
     extra_headers = kwargs.get("extra_headers", None)
@@ -4673,6 +4678,22 @@ def transcription(
         drop_params=drop_params,
     )
 
+    litellm_logging_obj.update_environment_variables(
+        model=model,
+        user=user,
+        optional_params={},
+        litellm_params={
+            "litellm_call_id": litellm_call_id,
+            "proxy_server_request": proxy_server_request,
+            "model_info": model_info,
+            "metadata": metadata,
+            "preset_cache_key": None,
+            "stream_response": {},
+            **kwargs,
+        },
+        custom_llm_provider=custom_llm_provider,
+    )
+
     response: Optional[TranscriptionResponse] = None
     if custom_llm_provider == "azure":
         # azure configs

litellm/proxy/proxy_cli.py

@@ -125,7 +125,7 @@ def is_port_in_use(port):
 )
 @click.option(
     "--request_timeout",
-    default=6000,
+    default=None,
     type=int,
     help="Set timeout in seconds for completion calls",
 )

tests/local_testing/test_whisper.py

@@ -115,3 +115,22 @@ async def test_transcription_caching():
     assert response_3.text != response_2.text
 
     litellm.cache = None
+
+
+@pytest.mark.asyncio
+async def test_whisper_log_pre_call():
+    from litellm.litellm_core_utils.litellm_logging import Logging
+    from datetime import datetime
+    from unittest.mock import patch, MagicMock
+    from litellm.integrations.custom_logger import CustomLogger
+
+    custom_logger = CustomLogger()
+
+    litellm.callbacks = [custom_logger]
+
+    with patch.object(custom_logger, "log_pre_api_call") as mock_log_pre_call:
+        await litellm.atranscription(
+            model="whisper-1",
+            file=audio_file,
+        )
+        mock_log_pre_call.assert_called_once()