Update app.py

app.py

@@ -6,73 +6,60 @@ from llama_index.core.tools import FunctionTool
 from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
 
 # --- 1. SET UP THE LLM ---
-# Use a strong model that supports tool calling/logic
-# You need your Hugging Face Token here
+# Replace with your actual HF Token (Settings -> Access Tokens)
 HF_TOKEN = "your_hf_token_here" 
+
+# We use Qwen2.5-Coder because it is excellent at logic and tool-calling
 llm = HuggingFaceInferenceAPI(
     model_name="Qwen/Qwen2.5-Coder-32B-Instruct", 
     token=HF_TOKEN
 )
 
 # --- 2. DEFINE THE TOOLS ---
-# LlamaIndex uses docstrings and type hints to explain tools to the LLM.
-
 def get_tokyo_time() -> str:
     """Useful for when you need to know the current time in Tokyo, Japan."""
     tz = pytz.timezone('Asia/Tokyo')
     now = datetime.datetime.now(tz)
-    return f"The current time in Tokyo is {now.strftime('%Y-%m-%d %H:%M:%S')} JST"
+    return f"The current time in Tokyo is {now.strftime('%H:%M:%S')} (JST)."
+
+def get_local_time() -> str:
+    """Useful for when you need to know the current time on the server's local machine."""
+    now = datetime.datetime.now()
+    return f"The local server time is {now.strftime('%H:%M:%S')}."
 
 def calculate(expression: str) -> str:
-    """Useful for performing mathematical calculations. Input should be a math string like '2+2'."""
+    """Useful for math calculations. Input should be a math string like '12 * 5' or '100 / 4'."""
     try:
-        # Simple safety check
+        # Basic security check
         allowed_chars = "0123456789+-*/(). "
         if not all(char in allowed_chars for char in expression):
-            return "Error: Invalid characters."
+            return "Error: Invalid math characters."
         return str(eval(expression))
     except Exception as e:
-        return f"Calculation error: {str(e)}"
+        return f"Error: {str(e)}"
 
-# Define a knowledge base tool
-def search_knowledge_base(query: str) -> str:
-    """Useful for answering questions about AI, ML, LLM, RAG, and Hugging Face."""
-    knowledge = {
-        "ai": "Artificial Intelligence is machines performing human-like tasks.",
-        "ml": "Machine Learning is a subset of AI where algorithms learn from data.",
-        "llm": "Large Language Models are AI systems trained on vast text.",
-        "rag": "RAG combines document retrieval with AI generation.",
-        "hugging face": "A platform for sharing AI models and datasets."
-    }
-    query_lower = query.lower()
-    for key, value in knowledge.items():
-        if key in query_lower:
-            return f"{key.upper()}: {value}"
-    return "I don't have that specific info in my knowledge base, but I can try to answer using my general knowledge."
+# Wrap functions into LlamaIndex tools
+tools = [
+    FunctionTool.from_defaults(fn=get_tokyo_time),
+    FunctionTool.from_defaults(fn=get_local_time),
+    FunctionTool.from_defaults(fn=calculate)
+]
 
-# Wrap them as LlamaIndex Tools
-time_tool = FunctionTool.from_defaults(fn=get_tokyo_time)
-calc_tool = FunctionTool.from_defaults(fn=calculate)
-kb_tool = FunctionTool.from_defaults(fn=search_knowledge_base)
-
-# --- 3. INITIALIZE THE AGENT ---
-agent = ReActAgent.from_tools(
-    [time_tool, calc_tool, kb_tool], 
-    llm=llm, 
-    verbose=True # This lets you see the "Thought" process in the console
-)
+# --- 3. INITIALIZE THE REACT AGENT ---
+# This agent will "think" step-by-step to choose the right tool
+agent = ReActAgent.from_tools(tools, llm=llm, verbose=True)
 
-# --- 4. GRADIO INTERFACE ---
-def agent_chat(message, history):
-    # LlamaIndex agents handle chat history automatically if you use chat()
+# --- 4. GRADIO CHAT INTERFACE ---
+def chat_with_agent(message, history):
+    # The agent.chat function manages the conversation logic
     response = agent.chat(message)
     return str(response)
 
 demo = gr.ChatInterface(
-    fn=agent_chat,
-    title="🤖 LlamaIndex Tokyo Agent",
-    description="Ask me about the time in Tokyo, math, or AI definitions!",
-    examples=["What is the time in Tokyo?", "Calculate 152 * 4", "Explain what RAG is."]
+    fn=chat_with_agent,
+    title="🦙 LlamaIndex Tokyo Agent",
+    description="I am a LlamaIndex-powered agent. I can tell you Tokyo time and do math!",
+    examples=["What is the time in Tokyo?", "What is 15 * 15?", "What time is it locally?"]
 )
 
 if __name__ == "__main__":