exception_type work

litellm/__init__.py

@@ -38,6 +38,7 @@ cache: Optional[Cache] = None # cache object
 model_alias_map: Dict[str, str] = {}
 max_budget: float = 0.0 # set the max budget across all providers
 _current_cost = 0 # private variable, used if max budget is set 
+error_logs: Dict = {}
 #############################################
 
 def get_model_cost_map():

litellm/main.py

@@ -1046,7 +1046,7 @@ def completion(
     except Exception as e:
         ## Map to OpenAI Exception
         raise exception_type(
-            model=model, custom_llm_provider=custom_llm_provider, original_exception=e
+            model=model, custom_llm_provider=custom_llm_provider, original_exception=e, completion_kwargs=args,
         )
 
 

litellm/tests/test_completion.py

@@ -134,24 +134,24 @@ def test_completion_with_litellm_call_id():
 #         pytest.fail(f"Error occurred: {e}")
 
 # using Non TGI or conversational LLMs
-# def hf_test_completion():
+def hf_test_completion():
-#     try:
+    try:
-#         # litellm.set_verbose=True
+        # litellm.set_verbose=True
-#         user_message = "My name is Merve and my favorite"
+        user_message = "My name is Merve and my favorite"
-#         messages = [{ "content": user_message,"role": "user"}]
+        messages = [{ "content": user_message,"role": "user"}]
-#         response = completion(
+        response = completion(
-#             model="huggingface/roneneldan/TinyStories-3M", 
+            model="huggingface/roneneldan/TinyStories-3M", 
-#             messages=messages,
+            messages=messages,
-#             api_base="https://p69xlsj6rpno5drq.us-east-1.aws.endpoints.huggingface.cloud",
+            api_base="https://p69xlsj6rpno5drq.us-east-1.aws.endpoints.huggingface.cloud",
-#             task=None,
-#         )
-#         # Add any assertions here to check the response
-#         print(response)
 
-#     except Exception as e:
+        )
-#         pytest.fail(f"Error occurred: {e}")
+        # Add any assertions here to check the response
+        print(response)
 
-# hf_test_completion()
+    except Exception as e:
+        pytest.fail(f"Error occurred: {e}")
+
+hf_test_completion()
 
 def test_completion_cohere(): # commenting for now as the cohere endpoint is being flaky
     try:
@@ -352,7 +352,7 @@ def test_completion_azure():
     try:
         print("azure gpt-3.5 test\n\n")
         response = completion(
-            model="azure/chatgpt-v-2",
+            model="chatgpt-v-2",
             messages=messages,
         )
         # Add any assertions here to check the response

litellm/utils.py

@@ -198,6 +198,7 @@ class Logging:
     def pre_call(self, input, api_key, model=None, additional_args={}):
         # Log the exact input to the LLM API
         print_verbose(f"Logging Details Pre-API Call for call id {self.litellm_call_id}")
+        litellm.error_logs['PRE_CALL'] = locals()
         try:
             # print_verbose(f"logging pre call for model: {self.model} with call type: {self.call_type}")
             self.model_call_details["input"] = input
@@ -280,6 +281,7 @@ class Logging:
 
     def post_call(self, original_response, input=None, api_key=None,  additional_args={}):
         # Log the exact result from the LLM API, for streaming - log the type of response received
+        litellm.error_logs['POST_CALL'] = locals()
         try:
             self.model_call_details["input"] = input
             self.model_call_details["api_key"] = api_key
@@ -1870,9 +1872,40 @@ def get_model_list():
         )
 
 ####### EXCEPTION MAPPING ################
-def exception_type(model, original_exception, custom_llm_provider):
+def exception_type(
+        model, 
+        original_exception, 
+        custom_llm_provider,
+        completion_kwargs={},
+    ):
     global user_logger_fn, liteDebuggerClient
     exception_mapping_worked = False
+
+    litellm.error_logs['EXCEPTION'] = original_exception
+    litellm.error_logs['KWARGS'] = completion_kwargs
+
+    import urllib.parse
+    import json
+    for log_key in litellm.error_logs:
+        current_logs = litellm.error_logs[log_key]
+        if type(current_logs) == dict:
+            filtered_error_logs = {key: value for key, value in current_logs.items() if isinstance(value, (str, int, float, bool, list, dict))}
+            litellm.error_logs[log_key] = filtered_error_logs
+        else:
+            litellm.error_logs[log_key] = str(current_logs)
+
+    # Convert the filtered_error_logs dictionary to a JSON string
+    error_logs_json = json.dumps(litellm.error_logs)
+    # URL-encode the JSON data
+    encoded_data = urllib.parse.quote(error_logs_json)
+    print(encoded_data)
+    # Print the encoded data (this is what you can include in a URL)
+    print("\033[91m" + str(litellm.error_logs) + "\033[0m")
+
+    decoded_data = urllib.parse.unquote(encoded_data)
+
+    # Print the decoded data
+    print(decoded_data)
     try:
         if isinstance(original_exception, OriginalError):
             # Handle the OpenAIError