Update app.py

app.py

@@ -19,8 +19,10 @@ HF_TOKEN = os.getenv("HF_TOKEN")
 # PART 1: LLAMAINDEX AGENT
 # ==========================================
 li_llm = HuggingFaceInferenceAPI(
+    # Explicitly using a model-id that is supported by Inference Providers
     model_name="Qwen/Qwen2.5-7B-Instruct",
     token=HF_TOKEN,
+    provider="together"  # CRITICAL: Fixes the 404 by using Together AI backend
 )
 
 def get_tokyo_time() -> str:
@@ -30,7 +32,6 @@ def get_tokyo_time() -> str:
 
 li_tools = [FunctionTool.from_defaults(fn=get_tokyo_time)]
 
-# Using positional argument for tools list to avoid TypeError
 li_agent = AgentWorkflow.from_tools_or_functions(
     li_tools,
     llm=li_llm,
@@ -38,7 +39,6 @@ li_agent = AgentWorkflow.from_tools_or_functions(
 
 async def chat_llama(message, history):
     try:
-        # AgentWorkflow requires async execution
         result = await li_agent.run(user_msg=message)
         return str(result)
     except Exception as e:
@@ -49,7 +49,8 @@ async def chat_llama(message, history):
 # ==========================================
 smol_model = InferenceClientModel(
     model_id="Qwen/Qwen2.5-Coder-32B-Instruct",
-    token=HF_TOKEN
+    token=HF_TOKEN,
+    provider="together" # CRITICAL: Ensures the smolagents client also finds a valid provider
 )
 
 @tool
@@ -60,7 +61,6 @@ def weather_tool(location: str) -> str:
     """
     return f"The weather in {location} is currently sunny and 22°C."
 
-# Tool initialization happens here; ensure ddgs is in requirements.txt
 smol_agent = CodeAgent(
     model=smol_model,
     tools=[weather_tool, DuckDuckGoSearchTool()]
@@ -68,7 +68,6 @@ smol_agent = CodeAgent(
 
 def chat_smol(message, history):
     try:
-        # CodeAgent .run() is synchronous
         response = smol_agent.run(message)
         return str(response)
     except Exception as e: