phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-26 11:14:04 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(anthropic_adapter.py): support streaming requests for /v1/messages endpoint

Browse source 
Fixes https://github.com/BerriAI/litellm/issues/5011
This commit is contained in:
Krrish Dholakia 2024-08-03 20:16:19 -07:00
parent b3aa722ebb
commit 5810708c71
9 changed files with 425 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

75
litellm/proxy/proxy_server.py
View file

@ -2396,7 +2396,9 @@ async def async_data_generator(
user_api_key_dict=user_api_key_dict, response=chunk
)
chunk = chunk.model_dump_json(exclude_none=True, exclude_unset=True)
if isinstance(chunk, BaseModel):
chunk = chunk.model_dump_json(exclude_none=True, exclude_unset=True)
try:
yield f"data: {chunk}\n\n"
except Exception as e:
@ -2437,6 +2439,59 @@ async def async_data_generator(
yield f"data: {error_returned}\n\n"
async def async_data_generator_anthropic(
response, user_api_key_dict: UserAPIKeyAuth, request_data: dict
):
verbose_proxy_logger.debug("inside generator")
try:
start_time = time.time()
async for chunk in response:
verbose_proxy_logger.debug(
"async_data_generator: received streaming chunk - {}".format(chunk)
)
### CALL HOOKS ### - modify outgoing data
chunk = await proxy_logging_obj.async_post_call_streaming_hook(
user_api_key_dict=user_api_key_dict, response=chunk
)
event_type = chunk.get("type")
try:
yield f"event: {event_type}\ndata:{json.dumps(chunk)}\n\n"
except Exception as e:
yield f"event: {event_type}\ndata:{str(e)}\n\n"
except Exception as e:
verbose_proxy_logger.error(
"litellm.proxy.proxy_server.async_data_generator(): Exception occured - {}\n{}".format(
str(e), traceback.format_exc()
)
)
await proxy_logging_obj.post_call_failure_hook(
user_api_key_dict=user_api_key_dict,
original_exception=e,
request_data=request_data,
)
verbose_proxy_logger.debug(
f"\033[1;31mAn error occurred: {e}\n\n Debug this by setting `--debug`, e.g. `litellm --model gpt-3.5-turbo --debug`"
)
router_model_names = llm_router.model_names if llm_router is not None else []
if isinstance(e, HTTPException):
raise e
else:
error_traceback = traceback.format_exc()
error_msg = f"{str(e)}\n\n{error_traceback}"
proxy_exception = ProxyException(
message=getattr(e, "message", error_msg),
type=getattr(e, "type", "None"),
param=getattr(e, "param", "None"),
code=getattr(e, "status_code", 500),
)
error_returned = json.dumps({"error": proxy_exception.to_dict()})
yield f"data: {error_returned}\n\n"
def select_data_generator(
response, user_api_key_dict: UserAPIKeyAuth, request_data: dict
):
@ -5379,6 +5434,19 @@ async def anthropic_response(
)
)
if (
"stream" in data and data["stream"] is True
): # use generate_responses to stream responses
selected_data_generator = async_data_generator_anthropic(
response=response,
user_api_key_dict=user_api_key_dict,
request_data=data,
)
return StreamingResponse(
selected_data_generator,
media_type="text/event-stream",
)
verbose_proxy_logger.info("\nResponse from Litellm:\n{}".format(response))
return response
except RejectedRequestError as e:
@ -5425,11 +5493,10 @@ async def anthropic_response(
user_api_key_dict=user_api_key_dict, original_exception=e, request_data=data
)
verbose_proxy_logger.error(
"litellm.proxy.proxy_server.completion(): Exception occured - {}".format(
str(e)
"litellm.proxy.proxy_server.anthropic_response(): Exception occured - {}\n{}".format(
str(e), traceback.format_exc()
)
)
verbose_proxy_logger.debug(traceback.format_exc())
error_msg = f"{str(e)}"
raise ProxyException(
message=getattr(e, "message", error_msg),