Tested locally, completed tutorial locally.

.gitignore

@@ -0,0 +1 @@
+.env

app.py

@@ -1,44 +1,45 @@
+import os
 import gradio as gr
-from huggingface_hub import InferenceClient
-
+from huggingface_hub import login
+from smolagents import DuckDuckGoSearchTool, InferenceClientModel, CodeAgent
+from tools import best_city, ClassifierTool
+
+web_search_tool = DuckDuckGoSearchTool()
+classifier_tool = ClassifierTool()
+
+hf_token = os.environ.get('HF_TOKEN')
+if hf_token:
+    login(token=hf_token) 
+
+model = InferenceClientModel(model_id='Qwen/Qwen3-4B-Instruct-2507', token=hf_token)
+
+tools = [
+    web_search_tool,
+    classifier_tool,
+    best_city
+]
+
+my_aiagent = CodeAgent(
+    tools=tools,
+    # For the purpose of this tutorial, just have tools you integrated.
+    # Also by default when teh add_base_tools is set to true, it will integrate DuckDuckGo Search.
+    add_base_tools=False, 
+    model=model
+)
 
 def respond(
     message,
     history: list[dict[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-    hf_token: gr.OAuthToken,
+    system_message
 ):
-    """
-    For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-    """
-    client = InferenceClient(token=hf_token.token, model="openai/gpt-oss-20b")
-
-    messages = [{"role": "system", "content": system_message}]
-
-    messages.extend(history)
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        choices = message.choices
-        token = ""
-        if len(choices) and choices[0].delta.content:
-            token = choices[0].delta.content
-
-        response += token
-        yield response
+    full_prompt = f"{system_message}\n\nChat history:\n{history}\n\nUser: {message}"
+    response = my_aiagent.run(
+        full_prompt,
+        max_steps=5, 
+        stream=False,
+    )
 
+    yield response
 
 """
 For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
@@ -46,23 +47,10 @@ For information on how to customize the ChatInterface, peruse the gradio docs: h
 chatbot = gr.ChatInterface(
     respond,
     type="messages",
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    additional_inputs=[],
 )
 
 with gr.Blocks() as demo:
-    with gr.Sidebar():
-        gr.LoginButton()
     chatbot.render()
 
 

requirements.txt

@@ -0,0 +1,5 @@
+smolagents 
+torch 
+transformers 
+huggingface_hub 
+ddgs

tools.py

@@ -0,0 +1,41 @@
+# All the tools you exported
+from transformers import pipeline
+from smolagents import tool, Tool
+
+@tool
+def best_city(input:str) -> str:
+    """
+    Suggests a the best city regardless of country
+    Args:
+        input (str): Any prompt indicating to get the best city. Allowed values are:
+                        - Kuala Lumpar, Malaysia
+    """
+    return "Kuala Lumpar, Malaysia"
+
+
+class ClassifierTool(Tool):
+  name = "zero_shot_classifier_tool"
+  description = "Classifies a sequence into given labels to determine if it is about a location or city."
+  inputs = {
+      "text": {"type": "string", "description": "The sequence to classify."}
+  }
+  output_type = "string"
+
+  def __init__(self, *args, **kwargs):
+    super().__init__(*args, **kwargs)
+    # Perform heavy computations such as initializing pipeline
+    self.classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")
+    self.location_labels = ['Favorite City', "Location", "City", 'Favorite Location', 'Best City', 'Best Location']
+    self.candidate_labels = [*self.location_labels, 'Other']
+
+  def forward(self, text: str) -> str:
+    print(f"Before Response::")
+    response = classifier(text, self.candidate_labels, multi_label=True, hypothesis_template="This prompt is about {}.")
+    print(f"Respoonse::: {response}")
+    # Check if any location labels meets the requirement
+    for label, score in zip(response['labels'], response['scores']):
+      print(f"Label: {label}, Score: {score:.4f}")
+      if label != 'Other' and score > 0.7:
+        return f"Match found: {label} (Confidence: {score:.4f})"
+
+    return "No location match found."