diff --git a/dsLightRag/Config/Config.py b/dsLightRag/Config/Config.py
index b5ff471e..c20acad6 100644
--- a/dsLightRag/Config/Config.py
+++ b/dsLightRag/Config/Config.py
@@ -3,15 +3,15 @@ ALY_AK = 'LTAI5tE4tgpGcKWhbZg6C4bh'
 ALY_SK = 'oizcTOZ8izbGUouboC00RcmGE8vBQ1'
 
 # 大模型 【DeepSeek深度求索官方】训练时用这个
-LLM_API_KEY = "sk-44ae895eeb614aa1a9c6460579e322f1"
-LLM_BASE_URL = "https://api.deepseek.com"
-LLM_MODEL_NAME = "deepseek-chat"
+#LLM_API_KEY = "sk-44ae895eeb614aa1a9c6460579e322f1"
+#LLM_BASE_URL = "https://api.deepseek.com"
+#LLM_MODEL_NAME = "deepseek-chat"
 
 # # 阿里云提供的大模型服务 【阿里云在处理文字材料时，容易引发绿网拦截，导致数据上报异常】
-# LLM_API_KEY = "sk-f6da0c787eff4b0389e4ad03a35a911f"
-# LLM_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
-# LLM_MODEL_NAME = "qwen-plus" # 不要使用通义千问，会导致化学方程式不正确！
-# #LLM_MODEL_NAME = "deepseek-v3"
+LLM_API_KEY = "sk-f6da0c787eff4b0389e4ad03a35a911f"
+LLM_BASE_URL = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+LLM_MODEL_NAME = "qwen-plus" # 不要使用通义千问，会导致化学方程式不正确！
+#LLM_MODEL_NAME = "deepseek-v3"
 
 # 嵌入向量模型
 EMBED_MODEL_NAME = "BAAI/bge-m3"
diff --git a/dsLightRag/Config/__pycache__/Config.cpython-310.pyc b/dsLightRag/Config/__pycache__/Config.cpython-310.pyc
index c6d5e1ad..6f95c86a 100644
Binary files a/dsLightRag/Config/__pycache__/Config.cpython-310.pyc and b/dsLightRag/Config/__pycache__/Config.cpython-310.pyc differ
diff --git a/dsLightRag/Doc/T1、史校长资料/测试结果/100题测试结果.xlsx b/dsLightRag/Doc/T1、史校长资料/测试结果/100题测试结果.xlsx
index 88638885..ca90e7b0 100644
Binary files a/dsLightRag/Doc/T1、史校长资料/测试结果/100题测试结果.xlsx and b/dsLightRag/Doc/T1、史校长资料/测试结果/100题测试结果.xlsx differ
diff --git a/dsLightRag/Test/T2_AnswerFromRAG.py b/dsLightRag/Test/T2_AnswerFromRAG.py
index 10a93cd0..9c7081a8 100644
--- a/dsLightRag/Test/T2_AnswerFromRAG.py
+++ b/dsLightRag/Test/T2_AnswerFromRAG.py
@@ -31,19 +31,23 @@ def getAnswer(query_text):
         return None
 
 
+# 要处理的Excel
 excel_file = r'D:\dsWork\dsProject\dsLightRag\Doc\T1、史校长资料\测试结果\100题测试结果.xlsx'
 
 if __name__ == '__main__':
-    # query_text = '什么是数学的性质命题'
-    # print(getAnswer(query_text))
-
     wb = load_workbook(excel_file)
     ws = wb.active  # 获取第一个sheet
     # 读取C列第2行到101行的数据
     idx = 1
     for row in range(2, 102):  # 行号从2到101
-        cell_value = ws[f'C{row}'].value
-        logger.info("问题{}:{}".format(idx, cell_value))
+        query_text = ws[f'C{row}'].value
+        logger.info("问题{}:{}".format(idx, query_text))
+        # 知识库作答
+        rag_answer = getAnswer(query_text)
+        logger.info("成功完成问题{}:{}".format(idx, query_text))
+        # 回写到Excel表中
+        ws['O'+str(idx+1)] = rag_answer
         idx = idx + 1
-
+    # 保存修改
+    wb.save(excel_file)
     wb.close()
diff --git a/dsLightRag/Test/T3_LLM.py b/dsLightRag/Test/T3_LLM.py
new file mode 100644
index 00000000..147e4489
--- /dev/null
+++ b/dsLightRag/Test/T3_LLM.py
@@ -0,0 +1,6 @@
+from Util import LlmUtil
+
+if __name__ == '__main__':
+    query_text = '你是谁？'
+    response = LlmUtil.get_llm_response(query_text)
+    print(response)
diff --git a/dsLightRag/Test/T4_Score.py b/dsLightRag/Test/T4_Score.py
new file mode 100644
index 00000000..37eb21d1
--- /dev/null
+++ b/dsLightRag/Test/T4_Score.py
@@ -0,0 +1,35 @@
+from openpyxl import load_workbook
+
+from Util import LlmUtil
+
+# 要处理的Excel
+excel_file = r'D:\dsWork\dsProject\dsLightRag\Doc\T1、史校长资料\测试结果\100题测试结果.xlsx'
+
+if __name__ == '__main__':
+    prompt = """
+    请针对下面这两段文字【1】和【2】进行评分，最高分为10分，最低分为0分，每1分差别为一个档次。
+    具体评分标准为：
+    只要【1】中包含【2】完全相同或者含义相同的内容，即从【1】可以推理出【2】即视为正确，打10分；
+    【1】中完全没有【2】表达的内容即视为错误打0分；趋于中间的按照涵盖程度打分，分越高越相似。
+    请注意：结果展示时，请按照如下标准进行输出:
+    先展示分析过程，然后展示最终得分，【最终得分】：x分，x替换为具体的分数，范围从0到10的整数。
+    """
+    wb = load_workbook(excel_file)
+    ws = wb.active  # 获取第一个sheet
+
+    query_text = prompt + "\n"
+    query_text = query_text + "下面将提供两段文字：\n"
+    query_text = query_text + "【1】：\n"
+    O2_content = ws['O2'].value
+    query_text = query_text + O2_content + "\n"
+
+    query_text = query_text + "【2】：\n"
+    O2_content = ws['D2'].value
+    query_text = query_text + O2_content + "\n"
+
+    response = LlmUtil.get_llm_response(query_text)
+    print(response)
+    # 对比
+    # 【1】:O2,【2】:D2 内容
+    # 结果保存到 P2 (打分过程) Q2(通过正则表达式截取出的 【最终得分】)
+    wb.close()
diff --git a/dsLightRag/Util/LlmUtil.py b/dsLightRag/Util/LlmUtil.py
new file mode 100644
index 00000000..8a9d1856
--- /dev/null
+++ b/dsLightRag/Util/LlmUtil.py
@@ -0,0 +1,15 @@
+from openai import OpenAI
+from Config.Config import *
+
+def get_llm_response(query_text:str):
+    # 获取大模型的响应
+    client = OpenAI(
+        api_key=LLM_API_KEY,
+        base_url=LLM_BASE_URL,
+    )
+    completion = client.chat.completions.create(
+        model=LLM_MODEL_NAME,
+        messages=[{'role': 'system', 'content': 'You are a helpful assistant.'},
+                  {'role': 'user', 'content': query_text}]
+    )
+    return completion.choices[0].message.content
\ No newline at end of file
diff --git a/dsLightRag/Util/__pycache__/LlmUtil.cpython-310.pyc b/dsLightRag/Util/__pycache__/LlmUtil.cpython-310.pyc
new file mode 100644
index 00000000..3c65256c
Binary files /dev/null and b/dsLightRag/Util/__pycache__/LlmUtil.cpython-310.pyc differ