update app.py

app.py

@@ -2,64 +2,51 @@ import gradio as gr
 from langchain_huggingface import HuggingFacePipeline
 from langchain.chains import ConversationChain
 from langchain.memory import ConversationBufferMemory
-from langchain_community.tools import DuckDuckGoSearchRun
-from langchain_core.tools import Tool
 
-# Load a smaller, faster model (TinyLlama-1.1B-Chat, optimized for CPU)
+# Load lightweight model (distilgpt2, ~82M parameters, fast on CPU)
 llm = HuggingFacePipeline.from_model_id(
-    model_id="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
+    model_id="distilgpt2",
     task="text-generation",
     pipeline_kwargs={
-        "max_new_tokens": 150,  # Shorter responses for speed
+        "max_new_tokens": 80,  # Very short responses for speed
         "do_sample": True,
-        "temperature": 0.6,    # Slightly less creative for consistency
+        "temperature": 0.7,    # Balanced creativity
         "top_k": 40,
         "top_p": 0.9
-    },
-    model_kwargs={"trust_remote_code": True}
+    }
 )
 
 # Set up conversation memory
 memory = ConversationBufferMemory()
 
-# Optional: Web search tool (comment out if you don’t need it for faster responses)
-search = DuckDuckGoSearchRun()
-tools = [
-    Tool(
-        name="Web Search",
-        func=search.run,
-        description="Use for current events or facts. Input a search query."
-    )
-]
-
-# Create a simple conversation chain (faster than full agent)
+# Create a simple conversation chain
 conversation = ConversationChain(
     llm=llm,
     memory=memory,
-    verbose=False  # Disable logging for speed
+    verbose=False  # No logging for speed
 )
 
-# Function to handle chat and optional tool use
+# Chat function
 def chat_with_agent(message, history):
     try:
-        # Check if the query likely needs a web search (e.g., current events)
-        if any(keyword in message.lower() for keyword in ["latest", "news", "current", "today", "weather"]):
-            search_result = search.run(message)
-            prompt = f"User asked: {message}\nWeb search result: {search_result}\nAnswer based on this info."
-            response = conversation.predict(input=prompt)
-        else:
-            response = conversation.predict(input=message)
+        response = conversation.predict(input=message)
+        # Clean up response to avoid verbosity
+        response = response.strip().split("\n")[0][:200]  # Limit to 200 chars
+        return response
     except Exception as e:
-        response = f"Error: {str(e)}. Try rephrasing your question."
-    return response
+        return f"Oops, something went wrong: {str(e)}. Try again!"
 
 # Gradio chat interface
 iface = gr.ChatInterface(
     fn=chat_with_agent,
-    title="Fast Free AI Agent",
-    description="A lightweight conversational AI that remembers our talks and can search the web. Hosted free on Hugging Face Spaces. Responses should be faster (5-15 seconds).",
-    examples=["What's my name if I told you it's Alex?", "Tell me a quick joke.", "What's the latest AI news?"]
+    title="Your AI Chat Buddy",
+    description="I'm here to talk with you and remember our conversation! Ask me anything, and I'll respond in a few seconds. Hosted free on Hugging Face Spaces.",
+    examples=[
+        "My name is Alex. What's my name?",
+        "Tell me a quick joke."
+    ]
 )
 
+# Launch the app
 if __name__ == "__main__":
     iface.launch()