Trial and error.

app.py

@@ -46,7 +46,9 @@ class BasicAgent:
                 {"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
                 {"role": "user", "content": question}
             ]
-            
+
+            print(f"Messages prepared for model: {messages}")
+
             # Generate response
             if self.model and self.tokenizer:
                 # Local model generation
@@ -57,6 +59,8 @@ class BasicAgent:
                 )
                 model_inputs = self.tokenizer([text], return_tensors="pt").to(self.model.device)
 
+                print(f"Model inputs prepared: {model_inputs}")
+
                 generated_ids = self.model.generate(
                     **model_inputs,
                     max_new_tokens=512
@@ -65,6 +69,8 @@ class BasicAgent:
                     output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
                 ]
 
+                print(f"Generated IDs: {generated_ids}")
+
                 answer = self.tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
             else:
                 # Fallback to Inference API