Update app.py

app.py

@@ -14,147 +14,29 @@ openai_client = OpenAI(api_key=os.getenv('OPENAI_API_KEY'))
 groq_client = Groq(api_key=os.getenv('groq_key'))
 
 # 模型設定
+OPENAI_MODELS = ["gpt-3.5-turbo", "gpt-4", "gpt-4-turbo"]
+GROQ_MODELS = ["llama3-8b-8192", "gemma2-9b-it"]
+
 MODEL_CONFIGS = {
     "openai": {
-        "gpt-3.5-turbo": {
-            "max_tokens": 4096,
-            "temperature": 0.7
-        },
-        "gpt-4": {
-            "max_tokens": 8192,
-            "temperature": 0.7
-        },
-        "gpt-4-turbo": {
-            "max_tokens": 4096,
-            "temperature": 0.7
-        }
+        "gpt-3.5-turbo": {"max_tokens": 4096, "temperature": 0.7},
+        "gpt-4": {"max_tokens": 8192, "temperature": 0.7},
+        "gpt-4-turbo": {"max_tokens": 4096, "temperature": 0.7}
     },
     "groq": {
-        "llama3-8b-8192": {
-            "max_tokens": 4090,
-            "temperature": 0.7
-        },
-        "gemma2-9b-it": {
-            "max_tokens": 1024,
-            "temperature": 0.7
-        }
+        "llama3-8b-8192": {"max_tokens": 4090, "temperature": 0.7},
+        "gemma2-9b-it": {"max_tokens": 1024, "temperature": 0.7}
     }
 }
 
-def get_llm_response(prompt, provider, model):
-    if provider == "openai":
-        response = openai_client.chat.completions.create(
-            model=model,
-            messages=[
-                {"role": "system", "content": "你是一位資深的國文作文評閱委員，請依據提供的評分規準進行評分。"},
-                {"role": "user", "content": prompt}
-            ],
-            **MODEL_CONFIGS["openai"][model]
-        )
-        return response.choices[0].message.content
-    else:  # groq
-        completion = groq_client.chat.completions.create(
-            model=model,
-            messages=[
-                {"role": "system", "content": "你是一位資深的國文作文評閱委員，請依據提供的評分規準進行評分。"},
-                {"role": "user", "content": prompt}
-            ],
-            **MODEL_CONFIGS["groq"][model],
-            stream=False,
-            top_p=1,
-            stop=None
-        )
-        return completion.choices[0].message.content
-
-def evaluate_essay(message, additional_prompt, provider, model):
-    if not message.strip():
-        return [], gr.Markdown("### 請輸入作文內容進行評分")
-    
-    criteria = {
-        '題旨發揮': {'weight': 0.4, 'max_score': 40},
-        '資料掌握': {'weight': 0.2, 'max_score': 20},
-        '結構安排': {'weight': 0.2, 'max_score': 20},
-        '字句運用': {'weight': 0.2, 'max_score': 20}
-    }
-    
-    grade_scores = {
-        'A+': 95, 'A': 90, 'A-': 85,
-        'B+': 80, 'B': 75, 'B-': 70,
-        'C+': 65, 'C': 60, 'C-': 55,
-        '0': 0
-    }
-
-    try:
-        history = []
-        total_score = 0
-        history.append(("作文內容：", message))
-        history.append(("", f"正在使用 {provider} ({model}) 進行評分分析..."))
-        
-        all_feedback = {}
-        
-        for criterion, details in criteria.items():
-            prompt = f"""評估以下作文的{criterion}（權重{details['weight']*100}%）：
-
-作文內容：
-{message}
-
-{additional_prompt if additional_prompt else ''}
-
-請依三等九級制（A+、A、A-、B+、B、B-、C+、C、C-）評分，並提供簡短評語。
-如果是缺考、未作答、完全文不對題或作答內容完全照抄試題，請給予0分。
-
-請按以下格式回覆：
-等第：
-評語："""
-
-            result = get_llm_response(prompt, provider, model)
-            lines = result.lower().split('\n')
-            grade = '0'
-            comment = ""
-            
-            for line in lines:
-                if '等第：' in line or '等第:' in line:
-                    grade_text = line.split('：')[-1].strip().upper()
-                    if grade_text in grade_scores:
-                        grade = grade_text
-                elif '評語：' in line or '評語:' in line:
-                    comment = line.split('：')[-1].strip()
-            
-            weighted_score = (grade_scores[grade] / 100) * details['max_score']
-            total_score += weighted_score
-            
-            feedback = f"### {criterion}\n"
-            feedback += f"- **等第**：{grade}\n"
-            feedback += f"- **得分**：{weighted_score:.1f}/{details['max_score']}\n"
-            feedback += f"- **評語**：{comment}\n"
-            
-            all_feedback[criterion] = feedback
-
-        for criterion in criteria:
-            history.append(("", all_feedback[criterion]))
-        
-        total_score_display = f"""
-# 總評分結果
-## 使用模型：{provider} ({model})
-## 總分：{total_score:.1f}/100
-"""
-        
-        return history, gr.Markdown(total_score_display)
-        
-    except Exception as e:
-        return [("", f"評分過程發生錯誤：{str(e)}")], gr.Markdown("### ❌ 評分失敗")
+[... 其他函數保持不變 ...]
 
+# 模型選擇切換函數
 def update_model_choices(provider):
     if provider == "openai":
-        return {
-            "choices": list(MODEL_CONFIGS["openai"].keys()),
-            "value": "gpt-3.5-turbo"
-        }
+        return gr.Dropdown(choices=OPENAI_MODELS, value="gpt-3.5-turbo")
     else:
-        return {
-            "choices": list(MODEL_CONFIGS["groq"].keys()),
-            "value": "llama3-8b-8192"
-        }
+        return gr.Dropdown(choices=GROQ_MODELS, value="llama3-8b-8192")
 
 # 建立 Gradio 介面
 with gr.Blocks(title="國文作文自動評分系統") as demo:
@@ -190,7 +72,7 @@ with gr.Blocks(title="國文作文自動評分系統") as demo:
             value="openai"
         )
         model = gr.Dropdown(
-            choices=list(MODEL_CONFIGS["openai"].keys()),
+            choices=OPENAI_MODELS,
             label="選擇模型",
             value="gpt-3.5-turbo",
             interactive=True