phoenix/litellm
1
0
Fork 0
forked from phoenix/litellm-mirror
Code Pull requests Activity

(fix) async + stream +sync logging

Browse source
This commit is contained in:
ishaan-jaff 2023-12-15 18:31:36 +05:30
parent ae10299058
commit fa57a6d1f8
1 changed files with 9 additions and 14 deletions
Download patch file Download diff file Expand all files Collapse all files

23
litellm/utils.py
View file

@ -856,22 +856,17 @@ class Logging:
print_verbose(f"success callbacks: {litellm.success_callback}") print_verbose(f"success callbacks: {litellm.success_callback}")
## BUILD COMPLETE STREAMED RESPONSE ## BUILD COMPLETE STREAMED RESPONSE
complete_streaming_response = None complete_streaming_response = None
if self.stream == True and self.model_call_details.get("litellm_params", {}).get("acompletion", False) == False: # only call stream chunk builder if it's not acompletion() if self.stream and self.model_call_details.get("litellm_params", {}).get("acompletion", False) == False: # only call stream chunk builder if it's not acompletion()
# if it's acompletion == True, chunks are built/appended in async_success_handler
if result.choices[0].finish_reason is not None: # if it's the last chunk if result.choices[0].finish_reason is not None: # if it's the last chunk
streaming_chunks = self.streaming_chunks + [result] self.streaming_chunks.append(result)
complete_streaming_response = litellm.stream_chunk_builder(streaming_chunks, messages=self.model_call_details.get("messages", None)) # print_verbose(f"final set of received chunks: {self.streaming_chunks}")
else: try:
# this is a completion() call
if self.stream == True:
print_verbose("success callback - assembling complete streaming response")
if result.choices[0].finish_reason is not None: # if it's the last chunk
print_verbose(f"success callback - Got the very Last chunk. Assembling {self.streaming_chunks}")
self.streaming_chunks.append(result)
complete_streaming_response = litellm.stream_chunk_builder(self.streaming_chunks, messages=self.model_call_details.get("messages", None)) complete_streaming_response = litellm.stream_chunk_builder(self.streaming_chunks, messages=self.model_call_details.get("messages", None))
print_verbose(f"success callback - complete streamign response{complete_streaming_response}") except:
else: complete_streaming_response = None
self.streaming_chunks.append(result) else:
self.streaming_chunks.append(result)
if complete_streaming_response: if complete_streaming_response:
self.model_call_details["complete_streaming_response"] = complete_streaming_response self.model_call_details["complete_streaming_response"] = complete_streaming_response