Update app.py

app.py

@@ -36,8 +36,6 @@ def get_prompt(act):
     matching_prompt = prompts_df[prompts_df['act'] == act]['prompt'].values
     return matching_prompt[0] if len(matching_prompt) > 0 else None
 
-
-
 def respond(
     message,
     history: list[tuple[str, str]],
@@ -51,20 +49,12 @@ def respond(
 반드시 한글로 답변할것. 
     """
     
-    messages = [{"role": "system", "content": f"{system_prefix} {system_message}"}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    # 사용자 입력에 따른 프롬프트 선택
-    prompt = get_prompt(message)
-    if prompt:
-        message = prompt
-
-    messages.append({"role": "user", "content": message})
+    full_prompt = f"{system_prefix} {system_message}\n\n"
+    
+    for user, assistant in history:
+        full_prompt += f"Human: {user}\nAI: {assistant}\n"
+    
+    full_prompt += f"Human: {message}\nAI:"
 
     API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-70B-Instruct"
     headers = {"Authorization": f"Bearer {os.getenv('HF_TOKEN')}"}
@@ -75,12 +65,12 @@ def respond(
 
     try:
         payload = {
-            "inputs": messages,
+            "inputs": full_prompt,
             "parameters": {
                 "max_new_tokens": max_tokens,
                 "temperature": temperature,
                 "top_p": top_p,
-                "stream": False  # 스트리밍을 비활성화
+                "return_full_text": False
             },
         }
         raw_response = query(payload)
@@ -101,8 +91,6 @@ def respond(
 
     yield response
 
-# 나머지 코드는 그대로 유지
-
 demo = gr.ChatInterface(
     respond,
     title="AI Auto Paper",