From 6805be7d1426c8c6c881570a93b48bb06bdcd13e Mon Sep 17 00:00:00 2001
From: bigbrother666 <zeming.zhao@gmail.com>
Date: Thu, 9 May 2024 14:56:58 +0800
Subject: [PATCH] fix env sample key

---
 client/README.md                      |  4 ++--
 client/backend/llms/openai_wrapper.py | 34 +++++----------------------
 2 files changed, 8 insertions(+), 30 deletions(-)

diff --git a/client/README.md b/client/README.md
index 9fe010d..4c4ba36 100644
--- a/client/README.md
+++ b/client/README.md
@@ -8,8 +8,8 @@
 
 - PROJECT_DIR="xxxx" #项目缓存文件夹（相对于client文件夹的路径），如果不设定就直接放在repo下面了
 - WS_LOG="verbose"  #设定是否开始debug观察，调试阶段建议开始，尤其可以观察到每一步接口调用的原始请求和返回
-- LLM_API_BASE： #使用本地大模型推理服务使用（本地加载大模型）的 host:port, 不配置默认走http://localhost:8000
-- DASHSCOPE_API_KEY="YOUR_DASHSCOPE_API_KEY" #使用阿里灵积大模型推理服务使用
+- LLM_API_BASE： #使用兼容openaiSDK的LLM服务或者本地大模型推理使用（不配置默认走http://localhost:8000）
+- LLM_API_KEY="YOUR_DASHSCOPE_API_KEY" #大模型推理服务API KEY(注册参考最下)
 - ZHIPUAI_API_KEY= #使用智谱大模型接口使用（目前只会调用glm4，model参数没有意义）
 - VOLC_KEY='AK|SK' #使用火山云翻译api使用，格式为AK|SK
 - EMBEDDING_MODEL_PATH='' #embedding模型的地址，注意需要填写完整的绝对路径
diff --git a/client/backend/llms/openai_wrapper.py b/client/backend/llms/openai_wrapper.py
index 253c936..d924b4b 100644
--- a/client/backend/llms/openai_wrapper.py
+++ b/client/backend/llms/openai_wrapper.py
@@ -1,12 +1,10 @@
-'''
+"""
 除了openai外，很多大模型提供商也都使用openai的SDK，对于这一类可以统一使用本wrapper
 这里演示使用deepseek提供的DeepSeek-V2
-'''
+"""
 
-import random
 import os
 from openai import OpenAI
-import time
 
 
 token = os.environ.get('LLM_API_KEY', "")
@@ -25,32 +23,12 @@ def openai_llm(messages: list, model: str, logger=None, **kwargs) -> str:
         logger.debug(f'model: {model}')
         logger.debug(f'kwargs:\n {kwargs}')
 
-    response = client.chat.completions.create(messages=messages, model=model, **kwargs)
-
-    for i in range(2):
-        if response and response.choices:
-            break
+    try:
+        response = client.chat.completions.create(messages=messages, model=model, **kwargs)
 
+    except Exception as e:
         if logger:
-            logger.warning(f"request failed. code: {response}\nretrying...")
-        else:
-            print(f"request failed. code: {response}\nretrying...")
-
-        time.sleep(1 + i * 30)
-        kwargs['seed'] = random.randint(1, 10000)
-        response = client.chat.completions.create(
-            messages=messages,
-            model=model,
-            **kwargs
-        )
-
-    if not response or not response.choices:
-        if logger:
-            logger.warning(
-                f"request failed. code: {response}\nabort after multiple retries...")
-        else:
-            print(
-                f"request failed. code: {response}\naborted after multiple retries...")
+            logger.error(f'openai_llm error: {e}')
         return ''
 
     if logger: