Update app.py

import gradio as gr
from bs4 import BeautifulSoup
from markdownify import MarkdownConverter
from playwright.sync_api import sync_playwright


def md(soup, **options):
return MarkdownConverter(**options).convert_soup(soup)


def main_fn(url: str, check: list[str], request: gr.Request):
user_agent = request.headers["user-agent"]

with sync_playwright() as p:
browser = p.chromium.launch(
args=[
"--single-process",
"--no-zygote",
"--no-sandbox",
"--disable-gpu",
"--disable-dev-shm-usage",
"--headless=new",
]
)
context = browser.new_context(user_agent=user_agent)
page = context.new_page()

response = page.goto(url=url)
status = response.status
content = page.content()
title = page.title()

browser.close()

soup = BeautifulSoup(content, features="html.parser")

# Remove <script> e <style> corretamente
for tag in ["script", "style"]:
for t in soup.find_all(tag):
t.decompose()

# Converte conteúdo principal para Markdown
body = soup.find("body")
main = soup.find("main")

strip_tags = check # <- aplica o que o usuário selecionou

if main:
markdown = md(main, strip=strip_tags)
else:
markdown = md(body, strip=strip_tags)

return f"{title}\n======\n\n{markdown}"


demo = gr.Interface(
fn=main_fn,
title="URL para Markdown V2",
description="""
<div style="width: fit-content; margin: 0 auto;">
Este app acessa o HTML da URL informada e converte em Markdown.
Utiliza o Playwright, então funciona com páginas dinâmicas como React.
</div>
<div style="width: fit-content; margin: 0 auto;">
É possível ignorar links (<code><a></code>), imagens (<code><img></code>) e outros elementos.
</div>""",
inputs=[
gr.Text(label="URL", placeholder="https://*****"),
gr.CheckboxGroup(
label="Ignorar tags no Markdown gerado",
choices=["a", "img", "noscript"],
value=[],
),
],
outputs=[
gr.TextArea(label="Markdown gerado", show_copy_button=True)
],
allow_flagging="never",
)

demo.launch(server_name="0.0.0.0")

app.py

@@ -8,7 +8,7 @@ def md(soup, **options):
     return MarkdownConverter(**options).convert_soup(soup)
 
 
-def main_fn(url: str, check: list[int], request: gr.Request):
+def main_fn(url: str, check: list[str], request: gr.Request):
     user_agent = request.headers["user-agent"]
 
     with sync_playwright() as p:
@@ -43,7 +43,7 @@ def main_fn(url: str, check: list[int], request: gr.Request):
     body = soup.find("body")
     main = soup.find("main")
 
-    strip_tags = []  # Extração completa, sem ignorar imagens nem links
+    strip_tags = check  # <- aplica o que o usuário selecionou
 
     if main:
         markdown = md(main, strip=strip_tags)
@@ -62,12 +62,12 @@ demo = gr.Interface(
         Utiliza o Playwright, então funciona com páginas dinâmicas como React.
     </div>
     <div style="width: fit-content; margin: 0 auto;">
-        O conteúdo é convertido para Markdown de forma limpa e copiável.
+        É possível ignorar links (<code>&lt;a&gt;</code>), imagens (<code>&lt;img&gt;</code>) e outros elementos.
     </div>""",
     inputs=[
         gr.Text(label="URL", placeholder="https://*****"),
         gr.CheckboxGroup(
-            label="Ignorar tags (sem efeito - tudo será extraído)",
+            label="Ignorar tags no Markdown gerado",
             choices=["a", "img", "noscript"],
             value=[],
         ),
@@ -78,4 +78,4 @@ demo = gr.Interface(
     allow_flagging="never",
 )
 
-demo.launch(server_name="0.0.0.0")
+demo.launch(server_name="0.0.0.0")