Update app.py

app.py

@@ -32,27 +32,21 @@ def chat_fn(message, history, model, system_message, max_tokens, temperature, to
     response_generator = respond(message, history_list, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty)
     
     for messages in response_generator:
-        # 将历史消息直接转换为 Gradio Chatbot 格式
         chatbot_messages = []
-        for user_msg, assistant_msg in messages:
-            if user_msg:  # 添加用户消息
-                chatbot_messages.append([user_msg, False])
-            if assistant_msg:  # 添加助手消息
-                chatbot_messages.append([assistant_msg, True])
+        for msg in messages:
+            if isinstance(msg, tuple):  # 如果是旧格式的元组
+                user_msg, assistant_msg = msg
+                if user_msg:
+                    chatbot_messages.append({"role": "user", "content": user_msg})
+                if assistant_msg:
+                    chatbot_messages.append({"role": "assistant", "content": assistant_msg})
+            else:  # 如果已经是字典格式
+                chatbot_messages.append(msg)
         yield chatbot_messages, messages
 
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    model,
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    top_k,
-    repeat_penalty,
-):
+def respond(message, history, model, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty):
+    
     global llm
     global llm_model
     
@@ -117,8 +111,18 @@ def respond(
     for output in stream:
         outputs += output
         token_count += len(output.split())
-        current_history = history + [(message, outputs)]
-        yield current_history
+        current_messages = []
+        
+        # 添加历史消息
+        for h in history:
+            current_messages.append({"role": "user", "content": h[0]})
+            current_messages.append({"role": "assistant", "content": h[1]})
+        
+        # 添加当前对话
+        current_messages.append({"role": "user", "content": message})
+        current_messages.append({"role": "assistant", "content": outputs})
+        
+        yield current_messages
 
     end_time = time.time()
     latency = end_time - start_time