2.17 增加读文件pdf接口测试1

flask_app/general/llm/多线程提问.py

@@ -5,7 +5,7 @@ import time
 from dashscope import Assistants, Messages, Runs, Threads
 from llama_index.indices.managed.dashscope import DashScopeCloudRetriever
 from flask_app.general.llm.大模型通用函数 import read_txt_to_string
-from flask_app.general.llm.通义千问long import qianwen_long, upload_file
+from flask_app.general.llm.通义千问long import qianwen_long, upload_file, qianwen_long_stream
 from flask_app.general.llm.qianwen_plus import qianwen_plus
 
 def read_questions_from_file(file_path):
@@ -126,8 +126,8 @@ def llm_call(question, knowledge_name,file_id, result_queue, ans_index, llm_type
             result_queue.put((ans_index, (question, ans)))  # 在队列中添加索引 (question, ans)
         elif llm_type==2:
             # print(f"qianwen_long! question:{question}")
-            # qianwen_res,usage = qianwen_long(file_id,question)   #有bug
-            qianwen_res = qianwen_long(file_id, question,2,1,need_extra)
+            qianwen_res = qianwen_long_stream(file_id, question,2,1,need_extra)    #流式返回
+            # qianwen_res = qianwen_long(file_id, question,2,1,need_extra)         #非流式
             if not qianwen_res:
                 result_queue.put((ans_index, None))  # 如果为空字符串，直接返回 None
             else:

flask_app/general/llm/通义千问long.py

@@ -74,7 +74,7 @@ def qianwen_long(file_id, user_query, max_retries=2, backoff_factor=1.0, need_ex
                 ],
                 stream=False
             )
-            token_usage = completion.usage.completion_tokens
+            token_usage = completion.usage.completion_tokens    #生成的tokens，不算输入的
             # 如果调用成功，返回响应内容
             if need_extra:
                 return completion.choices[0].message.content, token_usage
@@ -142,7 +142,6 @@ def qianwen_long_stream(file_id, user_query, max_retries=2, backoff_factor=1.0,
     - 当 need_extra=False 时: 返回响应内容 (str)
     - 当 need_extra=True 时: 返回 (响应内容, token_usage)
     """
-    print("调用 qianwen-long stream...")
 
     client = OpenAI(
         api_key=os.getenv("DASHSCOPE_API_KEY"),
@@ -181,7 +180,7 @@ def qianwen_long_stream(file_id, user_query, max_retries=2, backoff_factor=1.0,
                 # 处理 usage 信息
                 usage = chunk_data.get('usage')
                 if usage is not None:
-                    completion_tokens = usage.get('completion_tokens', 0)
+                    completion_tokens = usage.get('completion_tokens', 0)     #生成的tokens，不算输入的
 
                 # 处理 choices 信息
                 choices = chunk_data.get('choices', [])

flask_app/货物标/提取采购需求main.py

@@ -29,7 +29,7 @@ def fetch_procurement_reqs(procurement_path, invalid_path):
         page_count = get_pdf_page_count(procurement_path)
 
         if page_count > 60:    #转换过多，消耗的tokens 以及 文档转换的费用过高！退而求其次用qianwen-long
-            tech_model_type = 2  # long
+            tech_model_type = 2  # long-stream
             busi_model_type = 3  # long-stream
             processed_filepath = ""
             # 如果procurement_path截取成功，则更新invalid_path