Update app.py

app.py

@@ -13,10 +13,11 @@ import ollama
 
 # Model from run.sh
 MODEL_ID_MAP = {
+    "(阿里千問)Qwen3-0.6B": 'ollama pull hf.co/unsloth/Qwen3-0.6B-GGUF:Q5_K_M',
+    "(阿里千問)Qwen3-1.7B": 'ollama pull hf.co/unsloth/Qwen3-1.7B-GGUF:Q5_K_M',
+    "(阿里千問)Qwen3-4B-Instruct-2507": 'hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q4_K_M',
     "(IBM)Granite3.3-2B": 'granite3.3:2b',
     "(Meta)Llama3.2-3B-Instruct": 'hf.co/bartowski/Llama-3.2-3B-Instruct-GGUF:Q4_K_M', # OK speed with CPU
-    "(阿里千問)Qwen3-4B-Instruct-2507": 'hf.co/bartowski/Qwen_Qwen3-4B-Instruct-2507-GGUF:Q4_K_M',
-    "(阿里千問)Qwen3-Coder-30B-A3B-Instruct-1M": 'hf.co/unsloth/Qwen3-Coder-30B-A3B-Instruct-1M-GGUF:Q4_K_M',
     #"(阿里千問)Qwen3-4B-Thinking-2507": 'hf.co/bartowski/Qwen_Qwen3-4B-Thinking-2507-GGUF:Q4_K_M',
     #"(Google)Gemma3n-e2b-it": 'gemma3n:e2b-it-q4_K_M',
     #"(Tencent)混元-1.8B-Instruct":'hf.co/bartowski/tencent_Hunyuan-1.8B-Instruct-GGUF:Q4_K_M',
@@ -27,9 +28,8 @@ MODEL_ID_MAP = {
 
 # Default System Prompt
 DEFAULT_SYSTEM_PROMPT = """Answer everything in simple, smart, relevant and accurate style. No chatty! Besides, pls:
-    1. 如果查詢是以中文輸入，使用標準繁體中文回答，符合官方文書規範 
-    2. 要提供引用規則依据
-    3. 如果查詢是以英文輸入，使用英文回答"""
+    1. 如果查詢是以中文輸入，使用標準繁體中文回答
+    2. 如果查詢是以英文輸入，使用英文回答"""
 
 # --- Gradio Interface ---
 with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="neutral")) as demo:
@@ -78,15 +78,19 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="neutra
 
         # --- New: System Prompt Options ---
         SYSTEM_PROMPT_OPTIONS = {
-            "Smart & Accurate (Auto TC/EN)": DEFAULT_SYSTEM_PROMPT,
+            "Smart & Accurate & Auto TC/EN": DEFAULT_SYSTEM_PROMPT,
             "繁體中文回答":"無論如何，必須使用標準繁體中文回答. Answer everything in simple, smart, relevant and accurate style. No chatty!",
             "简体中文回答":"无论如何，必须使用标准简体中文回答. Answer everything in simple, smart, relevant and accurate style. No chatty!",
-            "English Caht":"You must reply by English. Answer everything in simple, smart, relevant and accurate style. No chatty!",
+            "English Chat":"You must reply by English. Answer everything in simple, smart, relevant and accurate style. No chatty!",
             "Friendly & Conversational":"Respond in a warm, friendly, and engaging tone. Use natural language and offer helpful suggestions. Keep responses concise but personable.",
             "Professional & Formal":"Maintain a formal and professional tone. Use precise language, avoid slang, and ensure responses are suitable for business or academic contexts.",
             "Elon Musk style":"You must chat in Elon Musk style!",
-            "Test2(Auto TC/EN)":"Always detect the user's input language and respond in that same language. Do not translate unless explicitly requested. Answer everything in simple, smart, relevant and accurate style. No chatty!",
-            "Coder":"""
+            "Test1(Auto TC/EN)":"Always detect the user's input language and respond in that same language. Do not translate unless explicitly requested. Answer everything in simple, smart, relevant and accurate style. No chatty!",
+            "Test2(Auto TC/EN)":"""Answer everything in simple, smart, relevant and accurate style. No chatty! Besides, pls:
+                1. 如果查詢是以中文輸入，使用標準繁體中文回答，符合官方文書規範 
+                2. 要提供引用規則依据
+                3. 如果查詢是以英文輸入，使用英文回答"""
+            "Good Coder":"""
                 You are a highly capable coding assistant specialized in software development, algorithms, and debugging. Your responses must be accurate, efficient, and tailored to the user's request. Always follow these principles:
                 1. Use clear, well-commented code.
                 2. Prioritize readability and best practices.
@@ -157,6 +161,10 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="neutra
         It takes the history, prepends the system prompt, calls the Ollama API,
         and streams the response back to the chatbot.
         """
+        
+        # Use selected predefined prompt unless custom is enabled
+        if not use_custom_prompt:
+            system_prompt = SYSTEM_PROMPT_OPTIONS[selected_prompt_key]
 
         current_selected_model = MODEL_ID_MAP[selected_model_name]
 
@@ -164,10 +172,9 @@ with gr.Blocks(theme=gr.themes.Default(primary_hue="blue", secondary_hue="neutra
         #if "Qwen3".lower() in current_selected_model:
         #    system_prompt = system_prompt+" /no_think"
 
-        # Use selected predefined prompt unless custom is enabled
-        if not use_custom_prompt:
-            system_prompt = SYSTEM_PROMPT_OPTIONS[selected_prompt_key]
-
+        if any(substring in current_selected_model.lower() for substring in ["qwen3-0.6b", "qwen3-1.7b"]):
+            system_prompt = system_prompt+" /no_think"
+        
         # The 'history' variable from Gradio contains the entire conversation.
         # We prepend the system prompt to this history to form the final payload.
         messages = [{"role": "system", "content": system_prompt}] + history