phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-26 03:04:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

(chore) undo changes on ishaan branch

Browse source
This commit is contained in:
ishaan-jaff 2023-11-24 16:55:57 -08:00
parent 0383e32628
commit 1a4958eb14
1 changed files with 70 additions and 40 deletions
Download patch file Download diff file Expand all files Collapse all files

110
litellm/proxy/utils.py
View file

@ -1,41 +1,71 @@
# import litellm import litellm
# from litellm import ModelResponse from litellm import ModelResponse
# from proxy_server import update_verification_token_cost from proxy_server import llm_model_list
# from typing import Optional from typing import Optional
# from fastapi import HTTPException, status
# import asyncio
# def track_cost_callback( def track_cost_callback(
# kwargs, # kwargs to completion kwargs, # kwargs to completion
# completion_response: ModelResponse, # response from completion completion_response: ModelResponse, # response from completion
# start_time = None, start_time = None,
# end_time = None, # start/end time for completion end_time = None, # start/end time for completion
# ): ):
# try: try:
# # init logging config # init logging config
# api_key = kwargs["litellm_params"]["metadata"]["api_key"] print("in custom callback tracking cost", llm_model_list)
# # check if it has collected an entire stream response if "azure" in kwargs["model"]:
# if "complete_streaming_response" in kwargs: # for azure cost tracking, we check the provided model list in the config.yaml
# # for tracking streaming cost we pass the "messages" and the output_text to litellm.completion_cost # we need to map azure/chatgpt-deployment to -> azure/gpt-3.5-turbo
# completion_response=kwargs["complete_streaming_response"] pass
# input_text = kwargs["messages"] # check if it has collected an entire stream response
# output_text = completion_response["choices"][0]["message"]["content"] if "complete_streaming_response" in kwargs:
# response_cost = litellm.completion_cost( # for tracking streaming cost we pass the "messages" and the output_text to litellm.completion_cost
# model = kwargs["model"], completion_response=kwargs["complete_streaming_response"]
# messages = input_text, input_text = kwargs["messages"]
# completion=output_text output_text = completion_response["choices"][0]["message"]["content"]
# ) response_cost = litellm.completion_cost(
# print(f"LiteLLM Proxy: streaming response_cost: {response_cost} for api_key: {api_key}") model = kwargs["model"],
# # for non streaming responses messages = input_text,
# else: completion=output_text
# # we pass the completion_response obj )
# if kwargs["stream"] != True: print("streaming response_cost", response_cost)
# response_cost = litellm.completion_cost(completion_response=completion_response) # for non streaming responses
# print(f"\n LiteLLM Proxy: regular response_cost: {response_cost} for api_key: {api_key}") else:
# we pass the completion_response obj
# ########### write costs to DB api_key / cost map if kwargs["stream"] != True:
# asyncio.run( input_text = kwargs.get("messages", "")
# update_verification_token_cost(token=api_key, additional_cost=response_cost) if isinstance(input_text, list):
# ) input_text = "".join(m["content"] for m in input_text)
# except: response_cost = litellm.completion_cost(completion_response=completion_response, completion=input_text)
# pass print("regular response_cost", response_cost)
except:
pass
def update_prisma_database(token, response_cost):
try:
# Import your Prisma client
from your_prisma_module import prisma
# Fetch the existing cost for the given token
existing_cost = prisma.LiteLLM_VerificationToken.find_unique(
where={
"token": token
}
).cost
# Calculate the new cost by adding the existing cost and response_cost
new_cost = existing_cost + response_cost
# Update the cost column for the given token
prisma_liteLLM_VerificationToken = prisma.LiteLLM_VerificationToken.update(
where={
"token": token
},
data={
"cost": new_cost
}
)
print(f"Prisma database updated for token {token}. New cost: {new_cost}")
except Exception as e:
print(f"Error updating Prisma database: {e}")
pass