Trial and error.

app.py

@@ -19,24 +19,6 @@ class BasicAgent:
 
         print("Loading Qwen2.5-7B-Instruct model...")
         self.model_name = "Qwen/Qwen2.5-7B-Instruct"
-        
-        # Load model and tokenizer
-        #try:
-        #    self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
-        #    self.model = AutoModelForCausalLM.from_pretrained(
-        #        self.model_name,
-        #        torch_dtype="auto",
-        #        device_map="auto"
-        #    )
-        #    print(f"Successfully loaded {self.model_name}")
-        #except Exception as e:
-        #    print(f"Error loading model: {e}")
-        #    # Fallback to HuggingFace Inference API if local loading fails
-        #    print("Falling back to InferenceClient")
-        #    self.client = InferenceClient(model=self.model_name)
-        #    self.tokenizer = None
-        #    self.model = None
-        print("Falling back to InferenceClient")
         self.client = InferenceClient(model=self.model_name)
         self.tokenizer = None
         self.model = None
@@ -72,8 +54,9 @@ class BasicAgent:
 
                 answer = self.tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
             else:
-                # Fallback to Inference API
-                answer = self.client.chat(messages=messages)
+                # Fallback to Inference API - using the correct method call for chat completion
+                response = self.client.chat_completion(messages)
+                answer = response.choices[0].message.content
                 
             print(f"Agent generated response (first 50 chars): {answer[:50]}...")
             return answer