From efdd4b8b8e32fc26496b828981a306a0ee1efea5 Mon Sep 17 00:00:00 2001
From: magicyuan876 <317617749@qq.com>
Date: Fri, 6 Dec 2024 15:23:18 +0800
Subject: [PATCH] =?UTF-8?q?=E7=A7=BB=E9=99=A4kwargs=E4=B8=AD=E7=9A=84hashi?=
 =?UTF-8?q?ng=5Fkv=E5=8F=82=E6=95=B0=E5=8F=96=E4=B8=BA=E5=8F=98=E9=87=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 lightrag/llm.py | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/lightrag/llm.py b/lightrag/llm.py
index d147e416..09e9fd74 100644
--- a/lightrag/llm.py
+++ b/lightrag/llm.py
@@ -73,11 +73,12 @@ async def openai_complete_if_cache(
     messages.extend(history_messages)
     messages.append({"role": "user", "content": prompt})
 
+    hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
     # Handle cache
     mode = kwargs.pop("mode", "default")
     args_hash = compute_args_hash(model, messages)
     cached_response, quantized, min_val, max_val = await handle_cache(
-        kwargs.get("hashing_kv"), args_hash, prompt, mode
+        hashing_kv, args_hash, prompt, mode
     )
     if cached_response is not None:
         return cached_response
@@ -219,12 +220,12 @@ async def bedrock_complete_if_cache(
 
     # Add user prompt
     messages.append({"role": "user", "content": [{"text": prompt}]})
-
+    hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
     # Handle cache
     mode = kwargs.pop("mode", "default")
     args_hash = compute_args_hash(model, messages)
     cached_response, quantized, min_val, max_val = await handle_cache(
-        kwargs.get("hashing_kv"), args_hash, prompt, mode
+        hashing_kv, args_hash, prompt, mode
     )
     if cached_response is not None:
         return cached_response
@@ -250,12 +251,12 @@ async def bedrock_complete_if_cache(
             args["inferenceConfig"][inference_params_map.get(param, param)] = (
                 kwargs.pop(param)
             )
-
+    hashing_kv: BaseKVStorage = kwargs.pop("hashing_kv", None)
     # Handle cache
     mode = kwargs.pop("mode", "default")
     args_hash = compute_args_hash(model, messages)
     cached_response, quantized, min_val, max_val = await handle_cache(
-        kwargs.get("hashing_kv"), args_hash, prompt, mode
+        hashing_kv, args_hash, prompt, mode
     )
     if cached_response is not None:
         return cached_response