Update app.py

app.py

@@ -170,6 +170,60 @@ def respond_claude_qna(
     except Exception as e:
         return f"예상치 못한 오류가 발생했습니다: {str(e)}"
 
+#############################
+# [추가코드] - Llama-3.3-70B-Instruct / Llama-3.2-3B-Instruct 적용
+#############################
+
+def get_llama_client(model_choice: str):
+    """
+    선택된 Llama 모델에 맞춰 InferenceClient 생성.
+    토큰은 환경 변수에서 가져옴.
+    """
+    hf_token = os.getenv("HF_TOKEN")
+    if not hf_token:
+        raise ValueError("HuggingFace API 토큰이 필요합니다.")
+    
+    if model_choice == "Llama-3.3-70B-Instruct":
+        model_id = "Llama-3.3-70B-Instruct"
+    elif model_choice == "Llama-3.2-3B-Instruct":
+        model_id = "Llama-3.2-3B-Instruct"
+    else:
+        raise ValueError("유효하지 않은 모델 선택입니다.")
+    return InferenceClient(model_id, token=hf_token)
+
+def respond_llama_qna(
+    question: str,
+    system_message: str,
+    max_tokens: int,
+    temperature: float,
+    top_p: float,
+    model_choice: str
+):
+    """
+    선택된 Llama 모델을 이용해 한 번의 질문(question)에 대한 답변을 반환하는 함수.
+    """
+    try:
+        client = get_llama_client(model_choice)
+    except ValueError as e:
+        return f"오류: {str(e)}"
+    
+    messages = [
+        {"role": "system", "content": system_message},
+        {"role": "user", "content": question}
+    ]
+    
+    try:
+        response_full = client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            temperature=temperature,
+            top_p=top_p,
+        )
+        assistant_message = response_full.choices[0].message.content
+        return assistant_message
+    except Exception as e:
+        return f"오류가 발생했습니다: {str(e)}"
+
 #############################
 # [기본코드] UI 부분 - 수정/삭제 불가
 #############################
@@ -405,8 +459,66 @@ with gr.Blocks() as demo:
             outputs=deepseek_answer_output
         )
 
+    #################
+    # Llama 탭 (추가)
+    #################
+    with gr.Tab("Llama"):
+        # 라디오 버튼 추가: Llama-3.3-70B-Instruct (기본) / Llama-3.2-3B-Instruct
+        llama_model_radio = gr.Radio(
+            choices=["Llama-3.3-70B-Instruct", "Llama-3.2-3B-Instruct"],
+            label="모델 선택",
+            value="Llama-3.3-70B-Instruct"
+        )
+
+        llama_input1 = gr.Textbox(label="입력1", lines=1)
+        llama_input2 = gr.Textbox(label="입력2", lines=1)
+        llama_input3 = gr.Textbox(label="입력3", lines=1)
+        llama_input4 = gr.Textbox(label="입력4", lines=1)
+        llama_input5 = gr.Textbox(label="입력5", lines=1)
+
+        llama_answer_output = gr.Textbox(label="결과", lines=5, interactive=False)
+
+        with gr.Accordion("고급 설정 (Llama)", open=False):
+            llama_system_message = gr.Textbox(
+                value="""반드시 한글로 답변할 것.
+너는 최고의 비서이다.
+내가 요구하는 것을 최대한 자세하고 정확하게 답변하라.
+""",
+                label="System Message",
+                lines=3
+            )
+            llama_max_tokens = gr.Slider(minimum=100, maximum=10000, value=4000, step=100, label="Max Tokens")
+            llama_temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
+            llama_top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P")
+
+        llama_submit_button = gr.Button("전송")
+
+        def merge_and_call_llama(i1, i2, i3, i4, i5, sys_msg, mt, temp, top_p_, model_choice):
+            question = " ".join([i1, i2, i3, i4, i5])
+            return respond_llama_qna(
+                question=question,
+                system_message=sys_msg,
+                max_tokens=mt,
+                temperature=temp,
+                top_p=top_p_,
+                model_choice=model_choice
+            )
+
+        llama_submit_button.click(
+            fn=merge_and_call_llama,
+            inputs=[
+                llama_input1, llama_input2, llama_input3, llama_input4, llama_input5,
+                llama_system_message,
+                llama_max_tokens,
+                llama_temperature,
+                llama_top_p,
+                llama_model_radio  # 라디오 버튼 입력 추가
+            ],
+            outputs=llama_answer_output
+        )
+
 #############################
 # 메인 실행부
 #############################
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()