Update app.py

app.py

@@ -9,17 +9,16 @@ from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
 # 1. SETUP LLM
 hf_token = os.getenv("HF_TOKEN")
 
-# Create the LLM with the 404 fix
+# Fix: Change task to "conversational" as required by the provider
 llm = HuggingFaceInferenceAPI(
     model_name="Qwen/Qwen2.5-7B-Instruct", 
     token=hf_token,
-    task="text-generation",
+    task="conversational", 
     provider="together",
     is_function_calling_model=False 
 )
 
-
-# 2. DEFINE YOUR TOOLS
+# 2. DEFINE YOUR TOOLS (Must be defined before the Agent)
 def get_tokyo_time() -> str:
     """Returns the current time in Tokyo, Japan."""
     tz = pytz.timezone('Asia/Tokyo')
@@ -29,13 +28,13 @@ def multiply(a: float, b: float) -> float:
     """Multiplies two numbers (a and b) and returns the result."""
     return a * b
 
+# Wrap them in LlamaIndex Tool objects
 tools = [
     FunctionTool.from_defaults(fn=get_tokyo_time),
     FunctionTool.from_defaults(fn=multiply)
 ]
 
 # 3. THE "STABILITY" PROMPT
-# This prevents the AI from freezing by giving it a clear pattern to follow.
 RE_ACT_PROMPT = """You are a helpful assistant.
 For every query, you MUST follow this sequence:
 Thought: <your reasoning>
@@ -48,6 +47,7 @@ Answer: <your final response to the user>
 """
 
 # 4. CREATE THE AGENT
+# Now 'tools' and 'llm' are both correctly defined and configured
 agent = ReActAgent.from_tools(
     tools, 
     llm=llm, 
@@ -58,11 +58,9 @@ agent = ReActAgent.from_tools(
 # 5. GRADIO INTERFACE
 def chat(message, history):
     try:
-        # Use .chat() to maintain the conversation flow
         response = agent.chat(message)
         return str(response)
     except Exception as e:
-        # If it still fails, this will tell us WHY in the chat window
         return f"System Error: {str(e)}"
 
 gr.ChatInterface(chat, title="Unit 2: LlamaIndex Agent (Fixed)").launch()