Update app.py

app.py

@@ -10,7 +10,7 @@ from llama_index.core.agent import ReActAgent
 from llama_index.core.tools import FunctionTool
 from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
 
-# ============ 1. SMOLAGENTS TOOLS (Strict Format) ============
+# ============ 1. SMOLAGENTS TOOLS ============
 
 @tool
 def time_tool(timezone: str) -> str:
@@ -53,7 +53,8 @@ def weather_tool(location: str) -> str:
     try:
         geo_url = f"https://geocoding-api.open-meteo.com/v1/search?name={location}&count=1&language=en&format=json"
         geo_res = requests.get(geo_url).json()
-        if not geo_res.get('results'): return "Location not found."
+        if not geo_res.get('results'): 
+            return "Location not found."
         data = geo_res['results'][0]
         weather_url = f"https://api.open-meteo.com/v1/forecast?latitude={data['latitude']}&longitude={data['longitude']}&current=temperature_2m"
         w_res = requests.get(weather_url).json()
@@ -61,6 +62,8 @@ def weather_tool(location: str) -> str:
     except Exception as e:
         return f"Error: {str(e)}"
 
+# ============ SMOLAGENTS SETUP ============
+
 HF_TOKEN = os.getenv("HF_TOKEN")
 smol_model = InferenceClientModel(model_id="Qwen/Qwen2.5-Coder-32B-Instruct", token=HF_TOKEN)
 
@@ -70,19 +73,18 @@ smol_agent = CodeAgent(
     max_steps=5
 )
 
-# ============ 2. LLAMAINDEX SETUP (STABLE VERSION) ============
+# ============ 2. LLAMAINDEX SETUP ============
 
-# FIX 1: Use 32B for LlamaIndex too. 7B is too weak for "ReAct" reasoning.
-# FIX 2: Use task="text-generation" (Stable for HF Inference API)
+# Use 32B model for better reasoning
 llama_llm = HuggingFaceInferenceAPI(
     model_name="Qwen/Qwen2.5-Coder-32B-Instruct", 
     token=HF_TOKEN,
     task="text-generation",
-    timeout=60
+    timeout=120,  # Increased timeout
+    context_window=32768,  # Set context window
 )
 
-# FIX 3: ReActAgent needs very clear instructions to stay in "Agent Mode"
-# We define tools with explicit Type Hints and Docstrings
+# Define tools with clear docstrings
 def get_tokyo_time() -> str:
     """Returns the current date and time in Tokyo, Japan."""
     tz = pytz.timezone('Asia/Tokyo')
@@ -90,42 +92,96 @@ def get_tokyo_time() -> str:
 
 def multiply(a: float, b: float) -> float:
     """Multiplies two numbers (a and b) and returns the result."""
-    return str(a * b)
+    return a * b
+
+def get_current_weather(city: str) -> str:
+    """Gets the current weather for a given city."""
+    try:
+        geo_url = f"https://geocoding-api.open-meteo.com/v1/search?name={city}&count=1&language=en&format=json"
+        geo_res = requests.get(geo_url).json()
+        if not geo_res.get('results'): 
+            return f"Could not find location: {city}"
+        data = geo_res['results'][0]
+        weather_url = f"https://api.open-meteo.com/v1/forecast?latitude={data['latitude']}&longitude={data['longitude']}&current=temperature_2m"
+        w_res = requests.get(weather_url).json()
+        return f"Weather in {data['name']}: {w_res['current']['temperature_2m']}°C"
+    except Exception as e:
+        return f"Weather API Error: {str(e)}"
 
 # Create the tools
 tools = [
     FunctionTool.from_defaults(fn=get_tokyo_time),
-    FunctionTool.from_defaults(fn=multiply)
+    FunctionTool.from_defaults(fn=multiply),
+    FunctionTool.from_defaults(fn=get_current_weather)
 ]
 
-# Initialize Agent with a clear "React" header
+# Initialize ReAct Agent with system prompt
 llama_agent = ReActAgent.from_tools(
     tools,
     llm=llama_llm,
-    verbose=True # This will print the "Thought/Action" in your terminal/logs
+    verbose=True,
+    max_iterations=10  # Allow more steps for complex queries
 )
 
-# ============ GRADIO UI (STABLE) ============
+# ============ CHAT FUNCTIONS ============
+
+def chat_smol(message, history):
+    """Handler for Smolagents chat"""
+    try:
+        response = smol_agent.run(message)
+        return str(response)
+    except Exception as e:
+        return f"⚠️ Smolagents Error: {str(e)}\n\nPlease try rephrasing your question."
 
 def chat_llama(message, history):
+    """Handler for LlamaIndex chat"""
     try:
-        # ReAct agents can sometimes take 2-3 "steps" to answer
+        # Chat method returns a response object
         response = llama_agent.chat(message)
         return str(response)
+    except AttributeError as e:
+        # Handle potential API issues
+        return f"⚠️ LlamaIndex API Error: {str(e)}\n\nTry: 'What is 5 times 7?' or 'What time is it in Tokyo?'"
     except Exception as e:
-        # This will tell you EXACTLY why it failed in the Chat window
-        return f"⚠️ LlamaIndex Error: {str(e)}"
+        return f"⚠️ LlamaIndex Error: {str(e)}\n\nThe agent may need a clearer question."
+
+# ============ GRADIO UI ============
 
 with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown("# 🤖 Multi-Framework AI Agent Space")
+    gr.Markdown("Compare two different agentic frameworks using the same LLM")
     
-    with gr.Tab("Smolagents (Alfred)"):
-        gr.Markdown("Uses `Qwen2.5-Coder-32B`")
-        gr.ChatInterface(chat_smol)
+    with gr.Tab("🔧 Smolagents (Alfred)"):
+        gr.Markdown("""
+        **Model:** Qwen2.5-Coder-32B-Instruct  
+        **Tools:** Web Search, Time, Weather, Translation
+        
+        Try: "What's the weather in Paris?" or "Translate 'hello' to Japanese"
+        """)
+        gr.ChatInterface(
+            chat_smol,
+            examples=[
+                "What's the weather in Tokyo?",
+                "What time is it in New York?",
+                "Translate 'good morning' to Spanish"
+            ]
+        )
+        
+    with gr.Tab("🧠 LlamaIndex (ReAct)"):
+        gr.Markdown("""
+        **Model:** Qwen2.5-Coder-32B-Instruct  
+        **Tools:** Tokyo Time, Multiply, Weather
         
-    with gr.Tab("LlamaIndex (ReAct)"):
-        gr.Markdown("Uses `Qwen2.5-7B`")
-        gr.ChatInterface(chat_llama)
+        Try: "What is 8 times 12?" or "What time is it in Tokyo?"
+        """)
+        gr.ChatInterface(
+            chat_llama,
+            examples=[
+                "What is 15 times 23?",
+                "What time is it in Tokyo?",
+                "What's the weather in London?"
+            ]
+        )
 
 if __name__ == "__main__":
     demo.launch()