From 2f3765a03f5ff659c8292edf3e0491297e477fe7 Mon Sep 17 00:00:00 2001 From: ishaan-jaff Date: Wed, 24 Jan 2024 18:51:39 -0800 Subject: [PATCH] (fix) log cache hits on SpendLogs table --- litellm/proxy/proxy_server.py | 6 ++++++ litellm/proxy/utils.py | 4 ++++ 2 files changed, 10 insertions(+) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index eaa2373cc34..e1ca25e130a 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -593,6 +593,12 @@ async def track_cost_callback( "user_api_key_user_id", None ) + if kwargs.get("cache_hit", False) == True: + response_cost = 0.0 + verbose_proxy_logger.info( + f"Cache Hit: response_cost {response_cost}, for user_id {user_id}" + ) + verbose_proxy_logger.info( f"response_cost {response_cost}, for user_id {user_id}" ) diff --git a/litellm/proxy/utils.py b/litellm/proxy/utils.py index 812157ca0e2..25c5c82ce53 100644 --- a/litellm/proxy/utils.py +++ b/litellm/proxy/utils.py @@ -1003,6 +1003,10 @@ def get_logging_payload(kwargs, response_obj, start_time, end_time): cache_key = litellm.cache.get_cache_key(**kwargs) else: cache_key = "Cache OFF" + if cache_hit == True: + import time + + id = f"{id}_cache_hit{time.time()}" # SpendLogs does not allow duplicate request_id payload = { "request_id": id,