OK, try this

app.py

@@ -35,13 +35,14 @@ Analyze the following IDA Hex Rays pseudocode and generate a valid JSON object c
 
 @spaces.GPU
 def predict(code):
+    prompt = """{system}
+    [INST]
+    {code}
+    [/INST]
+    """
     # Prepare messages with system prompt
-    messages = [
-        {"role": "system", "content": system},
-        {"role": "user", "content": "Hello, how are you?"}
-    ]
-    print(f"Test Tokenize: {tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)}")
-    return pipe(code)
+    #print(f"Test Tokenize: {tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)}")
+    return pipe(prompt)
 
 demo = gr.Interface(fn=predict, inputs="text", outputs="text",
                     description=frontmatter.load("README.md").content)