update

app.py

@@ -4,12 +4,14 @@ import os
 import gradio as gr
 import pandas as pd
 import requests
+from html_to_markdown import convert_to_markdown
 from smolagents import (
     CodeAgent,
     DuckDuckGoSearchTool,
     InferenceClientModel,
     load_tool,
     tool,
+    Tool,
 )
 
 
@@ -18,16 +20,22 @@ from smolagents import (
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 
+@tool
+def html_to_markdown(arg1: str) -> str:  # it's important to specify the return type
+    """A tool that converts HTML to Markdown.
+    Args:
+        arg1: the raw HTML string to convert
+    """
+    return convert_to_markdown(arg1)
+
+
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 class BasicAgent:
     def __init__(self):
         print("BasicAgent initialized.")
-        html_parser_tool = load_tool(
-            "victor/dom-to-semantic-markdown", trust_remote_code=True
-        )
         self.agent = CodeAgent(
-            tools=[html_parser_tool],
+            tools=[],
             model=InferenceClientModel(),
             additional_authorized_imports=["requests", "bs4"],
             max_steps=10,
@@ -36,8 +44,9 @@ class BasicAgent:
     def __call__(self, question: str) -> str:
         print(f"Agent received question: {question}")
         prompt = f"""
-            Answer the following question: {question}. Please follow the following rules:
-            1. When there is need to parse HTML please use LLM to extract the relevant information instead of using BeautifulSoup.
+            Answer the following question: `{question}`. 
+            Please follow the following rules:
+            1. When there is need to extract information from web page please use `html_to_markdown` tool to parse HTML content first, and then extract the information you need.
         """
         result = self.agent.run(prompt)
         print(f"Agent responded with: {result}")

requirements.txt

@@ -2,4 +2,5 @@ gradio
 requests
 smolagents
 duckduckgo_search
-bs4
+bs4
+html-to-markdown