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

fix(parallel_request_limiter.py): fix user+team tpm/rpm limit check

Browse source 
Closes https://github.com/BerriAI/litellm/issues/3788
This commit is contained in:
Krrish Dholakia 2024-05-27 08:48:23 -07:00
parent 3397acdf00
commit 56fd0c60d1
7 changed files with 157 additions and 532 deletions
Download patch file Download diff file Expand all files Collapse all files

4
litellm/proxy/utils.py
View file

@ -35,7 +35,6 @@ from litellm import (
)
from litellm.utils import ModelResponseIterator
from litellm.proxy.hooks.max_budget_limiter import _PROXY_MaxBudgetLimiter
from litellm.proxy.hooks.tpm_rpm_limiter import _PROXY_MaxTPMRPMLimiter
from litellm.proxy.hooks.cache_control_check import _PROXY_CacheControlCheck
from litellm.integrations.custom_logger import CustomLogger
from litellm.proxy.db.base_client import CustomDB
@ -81,9 +80,6 @@ class ProxyLogging:
self.call_details["user_api_key_cache"] = user_api_key_cache
self.internal_usage_cache = DualCache()
self.max_parallel_request_limiter = _PROXY_MaxParallelRequestsHandler()
self.max_tpm_rpm_limiter = _PROXY_MaxTPMRPMLimiter(
internal_cache=self.internal_usage_cache
)
self.max_budget_limiter = _PROXY_MaxBudgetLimiter()
self.cache_control_check = _PROXY_CacheControlCheck()
self.alerting: Optional[List] = None