Update app.py

app.py

@@ -13,24 +13,33 @@ MODEL_ENDPOINTS = {
     "Qwen2.5-Coder-32B-Instruct": "https://api-inference.huggingface.co/models/Qwen/Qwen2.5-Coder-32B-Instruct",
 }
 
-def query_model(prompt, model_endpoint):
+# System prompts for each model
+SYSTEM_PROMPTS = {
+    "Qwen2.5-72B-Instruct": "System: You are a knowledgeable assistant for general inquiries.",
+    "Llama3.3-70B-Instruct": "System: You are a research expert assistant specialized in in-depth analysis.",
+    "Qwen2.5-Coder-32B-Instruct": "System: You are a coding expert who helps with code-related tasks.",
+}
+
+def query_model(prompt, model_endpoint, system_prompt):
     headers = {
         "Authorization": f"Bearer {HF_API_KEY}",
         "Content-Type": "application/json",
         "Accept": "application/json"
     }
-    # Format prompt for instruct-style generation
-    formatted_prompt = f"User: {prompt}\nAssistant:"
+    # Combine the system prompt with the user prompt
+    formatted_prompt = f"{system_prompt}\nUser: {prompt}\nAssistant:"
+    
     data = {
         "inputs": formatted_prompt,
         "parameters": {
             "max_new_tokens": 512,
-            "temperature": 0.7,
+            "temperature": 0.6,  # All models use a temperature of 0.6
         }
     }
+    
     response = requests.post(model_endpoint, headers=headers, json=data)
     
-    # For debugging purposes, you can uncomment the following line:
+    # Uncomment the following line to print the raw API response for debugging:
     # print("Raw response:", response.text)
     
     try:
@@ -38,12 +47,10 @@ def query_model(prompt, model_endpoint):
     except Exception:
         return f"Error: Unable to parse JSON. Response: {response.text}"
     
-    # If the API returns an error message, surface it.
     if isinstance(result, dict) and "error" in result:
         return f"Error: {result['error']}"
     
     try:
-        # Expecting a list of outputs with a "generated_text" field.
         return result[0].get("generated_text", "No generated_text found in response")
     except Exception:
         return f"Error: Unexpected response format: {json.dumps(result)}"
@@ -51,7 +58,8 @@ def query_model(prompt, model_endpoint):
 def chat_with_models(user_input, history):
     responses = []
     for model_name, endpoint in MODEL_ENDPOINTS.items():
-        model_response = query_model(user_input, endpoint)
+        system_prompt = SYSTEM_PROMPTS.get(model_name, "")
+        model_response = query_model(user_input, endpoint, system_prompt)
         responses.append(f"**{model_name}**: {model_response}")
     combined_answer = "\n\n".join(responses)
     history.append((user_input, combined_answer))