fix model output

model.py

@@ -24,9 +24,15 @@ def run(
         add_generation_prompt=False,
         return_tensors="pt"
     )
+    input_ids = tokenized_chat.to(model.device)
+    input_length = input_ids.shape[1]
 
-    outputs = model.generate(tokenized_chat.to(model.device), max_new_tokens=2048)
-    output_text = tokenizer.decode(outputs[0])
+    outputs = model.generate(input_ids, max_new_tokens=2048)
+    # 2. Fatiamos o tensor: pegamos do [input_length:] até o fim
+    # Isso isola apenas os tokens novos gerados
+    generated_tokens = outputs[0][input_length:]
+
+    output_text = tokenizer.decode(generated_tokens, skip_special_tokens=True)
     return output_text
 
 if __name__ == "__main__":