update 0722

app.py

@@ -39,13 +39,6 @@ tokenizer = AutoTokenizer.from_pretrained("bert-base-chinese")
 model = AutoModel.from_pretrained("bert-base-chinese")
 
 
-
-# 本地加载数据
-dataset = load_data(file_path='train_300.json', num_samples=300)
-keyword_lists = [item['content'] for item in dataset if 'content' in item]
-summary_lists = [item['summary'] for item in dataset if 'summary' in item]
-
-
 global_api_key = None
 client = None
 
@@ -53,10 +46,6 @@ def initialize_clients(api_key):
     global client
     client = OpenAI(api_key=api_key)
 
-
-for item in keyword_lists:
-    item = item.split(',')
-
 def get_keywords(message):
     system_message = """
     # 角色
@@ -83,8 +72,7 @@ def get_keywords(message):
     return ','.join(keywords)
 
 
-
-
+#字符串匹配模块
 def keyword_match(query_keywords_dict, ad_keywords_lists, triggered_keywords, current_turn, window_size,distance_threshold):
     distance = 0
     most_matching_list = None
@@ -93,7 +81,6 @@ def keyword_match(query_keywords_dict, ad_keywords_lists, triggered_keywords, cu
     # query_keywords = query_keywords.split(',')
     # query_keywords = [keyword for keyword in query_keywords if keyword]
     
-    
     #匹配模块
     query_keywords= list(query_keywords_dict.keys())
     
@@ -126,6 +113,7 @@ def keyword_match(query_keywords_dict, ad_keywords_lists, triggered_keywords, cu
     
     return distance, index
 
+
 def encode_list_to_avg(keywords_list_list, model, tokenizer, device):
     if torch.cuda.is_available():
         print('Using GPU')
@@ -150,6 +138,7 @@ def encode_list_to_avg(keywords_list_list, model, tokenizer, device):
 
     return avg_embeddings
 
+
 def encode_to_avg(keywords_dict, model, tokenizer, device):
     if torch.cuda.is_available():
         print('Using GPU')
@@ -176,33 +165,6 @@ def encode_to_avg(keywords_dict, model, tokenizer, device):
     return avg_embedding.tolist()
 
 
-# def fetch_response_from_db(query_keywords,class_name):
-    
-#     avg_vec=np.array(encode_list_to_avg([query_keywords], model, tokenizer, device)[0])
-#     nearVector = {
-#         'vector': avg_vec
-#     }
-
-#     response = (
-#         db_client.query
-#         .get(class_name, ['keywords', 'summary'])
-#         .with_near_vector(nearVector)
-#         .with_limit(1)
-#         .with_additional(['distance'])
-#         .do()
-#     )
-
-#     print(response)
-#     class_name=class_name[0].upper()+class_name[1:]
-
-#     if class_name in response['data']['Get']:
-#         results = response['data']['Get'][class_name]
-#         return results[0]['_additional']['distance'],results[0]['summary'], results[0]['keywords']
-#     else:
-#         print(f"Class name {class_name} not found in response")
-#         return None
-
-
 def fetch_response_from_db(query_keywords_dict,class_name):
   
     
@@ -230,13 +192,15 @@ def fetch_response_from_db(query_keywords_dict,class_name):
         print(f"Class name {class_name} not found in response")
         return None
 
-def wrapper(message, history, max_tokens, temperature, top_p, window_size, distance_threshold, weight_keywords_users, weight_keywords_triggered, api_key):
+
+
+def wrapper(message, history, max_tokens, temperature, top_p, window_size, distance_threshold, weight_keywords_users, weight_keywords_triggered, api_key, state):
     initialize_clients(api_key) 
-    return respond(message, history, max_tokens, temperature, top_p, window_size, distance_threshold, weight_keywords_users, weight_keywords_triggered)
+    return respond(message, history, max_tokens, temperature, top_p, window_size, distance_threshold, weight_keywords_users, weight_keywords_triggered, state)
 
 
 #触发词及触发回合字典
-triggered_keywords = {}
+
 
 def respond(
     message,
@@ -247,10 +211,11 @@ def respond(
     window_size,
     distance_threshold,
     weight_keywords_users,
-    weight_keywords_triggered
+    weight_keywords_triggered,
+    triggered_keywords
 ):
 
-
+    triggered_keywords=triggered_keywords or {}
     system_message_with_ad = """
     # 角色
     你是一个热情的聊天机器人
@@ -291,9 +256,9 @@ def respond(
     keywords_dict={}
     for keywords in key_words_users:
         if keywords in keywords_dict:
-            keywords_dict[keywords]+=2
+            keywords_dict[keywords]+=weight_keywords_users
         else:
-            keywords_dict[keywords]=2
+            keywords_dict[keywords]=weight_keywords_users
     for keywords in key_words_assistant:
         if keywords in keywords_dict:
             keywords_dict[keywords]+=1
@@ -355,7 +320,7 @@ def respond(
         top_p=top_p,
     )
     
-    return response.choices[0].message.content
+    return response.choices[0].message.content , triggered_keywords
 
 
 # def chat_interface(message, history, max_tokens, temperature, top_p, window_size, distance_threshold):
@@ -388,7 +353,8 @@ demo = gr.ChatInterface(
         gr.Slider(minimum=0.01, maximum=0.20, value=0.08, step=0.01, label="Distance threshold"),
         gr.Slider(minimum=1, maximum=5, value=2, step=1, label="Weight of keywords from users"),
         gr.Slider(minimum=0, maximum=2, value=0.5, step=0.5, label="Weight of triggered keywords"),
-        gr.Textbox(label="api_key")
+        gr.Textbox(label="api_key"),
+        'state'
     ],
 )