Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,23 +1,31 @@
|
|
| 1 |
import os
|
| 2 |
os.system("playwright install")
|
| 3 |
-
# app.py (Final
|
| 4 |
|
| 5 |
import gradio as gr
|
| 6 |
-
from playwright.async_api import async_playwright
|
| 7 |
-
from bs4 import BeautifulSoup
|
| 8 |
import urllib.parse
|
| 9 |
import os
|
| 10 |
from itertools import cycle
|
| 11 |
import uuid
|
| 12 |
|
| 13 |
# --- 1. GLOBAL RESOURCES & STATE ---
|
| 14 |
-
# Initialized on the first request to be compatible with Spaces.
|
| 15 |
P = None
|
| 16 |
BROWSER = None
|
| 17 |
REVOLVER = None
|
| 18 |
LIVE_CONTEXTS = {}
|
| 19 |
APP_STARTED = False
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
# --- 2. PLAIN DATA STATE CLASSES ---
|
| 22 |
class TabState:
|
| 23 |
def __init__(self, tab_id, proxy_used="Direct Connection"):
|
|
@@ -44,16 +52,81 @@ class CredentialRevolver:
|
|
| 44 |
def get_next(self): return next(self.proxy_cycler) if self.proxy_cycler else None
|
| 45 |
def count(self): return len(self.proxies) if self.proxies else 0
|
| 46 |
|
| 47 |
-
# --- 3. CORE ASYNC LOGIC ---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
async def _fetch_and_update_tab_state(tab_state: TabState, url: str):
|
|
|
|
| 49 |
log = f"▶️ Navigating to {url}..."; live_page = LIVE_CONTEXTS[tab_state.id]["page"]
|
| 50 |
try:
|
| 51 |
await live_page.goto(url, wait_until='domcontentloaded', timeout=30000)
|
| 52 |
tab_state.url = live_page.url; tab_state.title = await live_page.title() or "No Title"
|
| 53 |
log += f"\n✅ Arrived at: {tab_state.url}"
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
tab_state.links = [{'text': link.get_text(strip=True) or "[No Link Text]", 'url': urllib.parse.urljoin(tab_state.url, link['href'])} for link in soup.find_all('a', href=True) if link.get('href', '').startswith('http')]
|
| 58 |
log += f"\n🔗 Found {len(tab_state.links)} links."
|
| 59 |
except Exception as e:
|
|
@@ -62,19 +135,28 @@ async def _fetch_and_update_tab_state(tab_state: TabState, url: str):
|
|
| 62 |
tab_state.links = []; log += f"\n❌ {error_message}"
|
| 63 |
return log
|
| 64 |
|
| 65 |
-
async def handle_action(browser_state: BrowserState, action: str, value=None):
|
|
|
|
| 66 |
log = ""; active_tab_state = browser_state.get_active_tab()
|
| 67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
tab_id, proxy_config = str(uuid.uuid4()), REVOLVER.get_next()
|
| 69 |
context = await BROWSER.new_context(proxy=proxy_config)
|
| 70 |
page = await context.new_page()
|
| 71 |
LIVE_CONTEXTS[tab_id] = {"context": context, "page": page}
|
| 72 |
new_tab = TabState(tab_id, proxy_used=proxy_config['server'].split('@')[-1] if proxy_config else "Direct")
|
| 73 |
browser_state.tabs.append(new_tab); browser_state.active_tab_id = tab_id
|
| 74 |
-
log = await _fetch_and_update_tab_state(new_tab, "https://www.
|
| 75 |
-
elif action == "go" and active_tab_state and value:
|
| 76 |
-
url = value if (urllib.parse.urlparse(value).scheme and urllib.parse.urlparse(value).netloc) else f"https://duckduckgo.com/html/?q={urllib.parse.quote_plus(value)}"
|
| 77 |
-
log = await _fetch_and_update_tab_state(active_tab_state, url)
|
| 78 |
elif action == "click" and active_tab_state and value is not None:
|
| 79 |
try:
|
| 80 |
link_index = int(value)
|
|
@@ -92,37 +174,37 @@ async def handle_action(browser_state: BrowserState, action: str, value=None):
|
|
| 92 |
browser_state.active_tab_id = value; log = f"Switched to tab."
|
| 93 |
return browser_state, log
|
| 94 |
|
| 95 |
-
def update_ui_components(browser_state: BrowserState):
|
| 96 |
-
active_tab = browser_state.get_active_tab()
|
| 97 |
-
if not active_tab: return {page_content: gr.Markdown("No active tabs."), url_textbox: "", links_display: "", tab_selector: gr.Radio(choices=[])}
|
| 98 |
-
tab_choices = [(f"Tab {i}: {t.title[:25]}... ({t.proxy_used})", t.id) for i, t in enumerate(browser_state.tabs)]
|
| 99 |
-
links_md = "### 🔗 Links on Page\n" + ('\n'.join(f"{i}. [{link['text'][:80]}]({link['url']})" for i, link in enumerate(active_tab.links[:25])) if active_tab.links else "_No links found._")
|
| 100 |
-
return {
|
| 101 |
-
page_content: gr.Markdown(f"# {active_tab.title}\n**URL:** {active_tab.url}\n\n---\n\n{active_tab.parsed_text[:2000]}..."),
|
| 102 |
-
url_textbox: gr.Textbox(value=active_tab.url), links_display: gr.Markdown(links_md),
|
| 103 |
-
tab_selector: gr.Radio(choices=tab_choices, value=active_tab.id, label="Active Tabs"),
|
| 104 |
-
}
|
| 105 |
-
|
| 106 |
# --- 4. GRADIO UI AND EVENT HANDLING ---
|
| 107 |
with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
| 108 |
browser_state = gr.State(BrowserState())
|
| 109 |
-
gr.Markdown("# 🛰️ Real Browser Demo
|
| 110 |
|
| 111 |
with gr.Row():
|
| 112 |
-
with gr.Column(scale=
|
| 113 |
-
|
| 114 |
-
|
| 115 |
-
|
| 116 |
-
|
| 117 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 118 |
tab_selector = gr.Radio(choices=[], label="Active Tabs", interactive=True)
|
| 119 |
with gr.Accordion("Clickable Links", open=True):
|
| 120 |
links_display = gr.Markdown("...");
|
| 121 |
-
with gr.Row():
|
|
|
|
|
|
|
| 122 |
|
|
|
|
|
|
|
| 123 |
all_outputs = [page_content, url_textbox, links_display, tab_selector, log_display]
|
| 124 |
|
| 125 |
-
async def master_handler(current_state, action, value=None):
|
| 126 |
global APP_STARTED, P, BROWSER, REVOLVER
|
| 127 |
if not APP_STARTED:
|
| 128 |
print("🚀 First request received, starting up Playwright...");
|
|
@@ -130,8 +212,8 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
|
| 130 |
proxy_list_str = os.getenv("PROXY_LIST", ""); REVOLVER = CredentialRevolver(proxy_list_str)
|
| 131 |
print(f"✅ Playwright started. {REVOLVER.count()} proxies loaded."); APP_STARTED = True
|
| 132 |
|
| 133 |
-
new_state, log = await handle_action(current_state, action, value)
|
| 134 |
-
ui_updates = update_ui_components(
|
| 135 |
|
| 136 |
return (
|
| 137 |
new_state,
|
|
@@ -142,29 +224,33 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
|
| 142 |
log
|
| 143 |
)
|
| 144 |
|
| 145 |
-
#
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
|
| 153 |
-
|
| 154 |
-
|
| 155 |
-
|
| 156 |
-
|
| 157 |
-
|
| 158 |
-
return await master_handler(state, "close_tab", None)
|
| 159 |
-
async def on_switch_tab(state, value):
|
| 160 |
-
return await master_handler(state, "switch_tab", value)
|
| 161 |
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 169 |
|
| 170 |
demo.launch()
|
|
|
|
| 1 |
import os
|
| 2 |
os.system("playwright install")
|
| 3 |
+
# app.py (Final Version with Search Engines and Improved Formatting)
|
| 4 |
|
| 5 |
import gradio as gr
|
| 6 |
+
from playwright.async_api import async_playwright
|
| 7 |
+
from bs4 import BeautifulSoup, NavigableString
|
| 8 |
import urllib.parse
|
| 9 |
import os
|
| 10 |
from itertools import cycle
|
| 11 |
import uuid
|
| 12 |
|
| 13 |
# --- 1. GLOBAL RESOURCES & STATE ---
|
|
|
|
| 14 |
P = None
|
| 15 |
BROWSER = None
|
| 16 |
REVOLVER = None
|
| 17 |
LIVE_CONTEXTS = {}
|
| 18 |
APP_STARTED = False
|
| 19 |
|
| 20 |
+
# New: Search Engine Configuration
|
| 21 |
+
SEARCH_ENGINES = {
|
| 22 |
+
"DuckDuckGo": "https://duckduckgo.com/html/?q={query}",
|
| 23 |
+
"Google": "https://www.google.com/search?q={query}",
|
| 24 |
+
"Bing": "https://www.bing.com/search?q={query}",
|
| 25 |
+
"Brave": "https://search.brave.com/search?q={query}",
|
| 26 |
+
"Ecosia": "https://www.ecosia.org/search?q={query}"
|
| 27 |
+
}
|
| 28 |
+
|
| 29 |
# --- 2. PLAIN DATA STATE CLASSES ---
|
| 30 |
class TabState:
|
| 31 |
def __init__(self, tab_id, proxy_used="Direct Connection"):
|
|
|
|
| 52 |
def get_next(self): return next(self.proxy_cycler) if self.proxy_cycler else None
|
| 53 |
def count(self): return len(self.proxies) if self.proxies else 0
|
| 54 |
|
| 55 |
+
# --- 3. CORE ASYNC LOGIC with NEW FORMATTER ---
|
| 56 |
+
|
| 57 |
+
def _format_html_to_markdown(soup):
|
| 58 |
+
"""
|
| 59 |
+
NEW: Intelligently converts a BeautifulSoup object to formatted Markdown.
|
| 60 |
+
"""
|
| 61 |
+
text_parts = []
|
| 62 |
+
|
| 63 |
+
# Use a more specific main content area if available
|
| 64 |
+
content_node = soup.find('main') or soup.find('body')
|
| 65 |
+
if not content_node:
|
| 66 |
+
return "Could not find main body content."
|
| 67 |
+
|
| 68 |
+
for element in content_node.find_all(recursive=False):
|
| 69 |
+
# Ignore common non-content sections
|
| 70 |
+
if element.name in ['nav', 'footer', 'header', 'aside', 'form', 'script', 'style']:
|
| 71 |
+
continue
|
| 72 |
+
text_parts.append(_process_element_to_markdown(element))
|
| 73 |
+
|
| 74 |
+
return '\n'.join(filter(None, text_parts))
|
| 75 |
+
|
| 76 |
+
def _process_element_to_markdown(element):
|
| 77 |
+
"""NEW: Recursive helper to process each element into Markdown."""
|
| 78 |
+
if isinstance(element, NavigableString):
|
| 79 |
+
return element.strip()
|
| 80 |
+
|
| 81 |
+
if element.name is None:
|
| 82 |
+
return ''
|
| 83 |
+
|
| 84 |
+
# Get the inner text by recursively processing children
|
| 85 |
+
inner_text = ''.join(_process_element_to_markdown(child) for child in element.children)
|
| 86 |
+
|
| 87 |
+
# Apply formatting based on tag name
|
| 88 |
+
if element.name in ['p', 'div', 'article', 'section']:
|
| 89 |
+
return f"\n{inner_text.strip()}\n"
|
| 90 |
+
if element.name == 'h1':
|
| 91 |
+
return f"\n# {inner_text.strip()}\n"
|
| 92 |
+
if element.name == 'h2':
|
| 93 |
+
return f"\n## {inner_text.strip()}\n"
|
| 94 |
+
if element.name == 'h3':
|
| 95 |
+
return f"\n### {inner_text.strip()}\n"
|
| 96 |
+
if element.name in ['h4', 'h5', 'h6']:
|
| 97 |
+
return f"\n#### {inner_text.strip()}\n"
|
| 98 |
+
if element.name == 'li':
|
| 99 |
+
return f"* {inner_text.strip()}\n"
|
| 100 |
+
if element.name in ['ul', 'ol']:
|
| 101 |
+
return f"\n{inner_text}\n"
|
| 102 |
+
if element.name in ['strong', 'b']:
|
| 103 |
+
return f"**{inner_text.strip()}**"
|
| 104 |
+
if element.name in ['em', 'i']:
|
| 105 |
+
return f"*{inner_text.strip()}*"
|
| 106 |
+
if element.name in ['pre', 'code']:
|
| 107 |
+
return f"\n```\n{inner_text.strip()}\n```\n"
|
| 108 |
+
if element.name == 'a':
|
| 109 |
+
href = element.get('href', '')
|
| 110 |
+
return f"[{inner_text.strip()}]({href})"
|
| 111 |
+
if element.name == 'hr':
|
| 112 |
+
return "\n---\n"
|
| 113 |
+
|
| 114 |
+
return inner_text # For other tags like span, etc.
|
| 115 |
+
|
| 116 |
async def _fetch_and_update_tab_state(tab_state: TabState, url: str):
|
| 117 |
+
"""Uses Playwright to navigate and the new formatter to parse."""
|
| 118 |
log = f"▶️ Navigating to {url}..."; live_page = LIVE_CONTEXTS[tab_state.id]["page"]
|
| 119 |
try:
|
| 120 |
await live_page.goto(url, wait_until='domcontentloaded', timeout=30000)
|
| 121 |
tab_state.url = live_page.url; tab_state.title = await live_page.title() or "No Title"
|
| 122 |
log += f"\n✅ Arrived at: {tab_state.url}"
|
| 123 |
+
|
| 124 |
+
html_content = await live_page.content()
|
| 125 |
+
soup = BeautifulSoup(html_content, 'lxml')
|
| 126 |
+
|
| 127 |
+
# Use the new formatter
|
| 128 |
+
tab_state.parsed_text = _format_html_to_markdown(soup)
|
| 129 |
+
|
| 130 |
tab_state.links = [{'text': link.get_text(strip=True) or "[No Link Text]", 'url': urllib.parse.urljoin(tab_state.url, link['href'])} for link in soup.find_all('a', href=True) if link.get('href', '').startswith('http')]
|
| 131 |
log += f"\n🔗 Found {len(tab_state.links)} links."
|
| 132 |
except Exception as e:
|
|
|
|
| 135 |
tab_state.links = []; log += f"\n❌ {error_message}"
|
| 136 |
return log
|
| 137 |
|
| 138 |
+
async def handle_action(browser_state: BrowserState, search_engine: str, action: str, value=None):
|
| 139 |
+
"""Main logic handler, now with search engine selection."""
|
| 140 |
log = ""; active_tab_state = browser_state.get_active_tab()
|
| 141 |
+
|
| 142 |
+
if action == "go" and active_tab_state and value:
|
| 143 |
+
is_url = urllib.parse.urlparse(value).scheme and urllib.parse.urlparse(value).netloc
|
| 144 |
+
if is_url:
|
| 145 |
+
url = value
|
| 146 |
+
else: # It's a search query
|
| 147 |
+
base_url = SEARCH_ENGINES.get(search_engine, SEARCH_ENGINES["DuckDuckGo"])
|
| 148 |
+
url = base_url.format(query=urllib.parse.quote_plus(value))
|
| 149 |
+
log = await _fetch_and_update_tab_state(active_tab_state, url)
|
| 150 |
+
|
| 151 |
+
# --- Other actions remain the same ---
|
| 152 |
+
elif action == "new_tab":
|
| 153 |
tab_id, proxy_config = str(uuid.uuid4()), REVOLVER.get_next()
|
| 154 |
context = await BROWSER.new_context(proxy=proxy_config)
|
| 155 |
page = await context.new_page()
|
| 156 |
LIVE_CONTEXTS[tab_id] = {"context": context, "page": page}
|
| 157 |
new_tab = TabState(tab_id, proxy_used=proxy_config['server'].split('@')[-1] if proxy_config else "Direct")
|
| 158 |
browser_state.tabs.append(new_tab); browser_state.active_tab_id = tab_id
|
| 159 |
+
log = await _fetch_and_update_tab_state(new_tab, "https://www.startpage.com/") # New default page
|
|
|
|
|
|
|
|
|
|
| 160 |
elif action == "click" and active_tab_state and value is not None:
|
| 161 |
try:
|
| 162 |
link_index = int(value)
|
|
|
|
| 174 |
browser_state.active_tab_id = value; log = f"Switched to tab."
|
| 175 |
return browser_state, log
|
| 176 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 177 |
# --- 4. GRADIO UI AND EVENT HANDLING ---
|
| 178 |
with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
| 179 |
browser_state = gr.State(BrowserState())
|
| 180 |
+
gr.Markdown("# 🛰️ Real Browser Demo v2")
|
| 181 |
|
| 182 |
with gr.Row():
|
| 183 |
+
with gr.Column(scale=4):
|
| 184 |
+
with gr.Row():
|
| 185 |
+
url_textbox = gr.Textbox(label="Enter URL or Search Query", interactive=True, scale=4)
|
| 186 |
+
go_btn = gr.Button("Go", variant="primary", scale=1)
|
| 187 |
+
with gr.Accordion("Page Content (Formatted)", open=True):
|
| 188 |
+
page_content = gr.Markdown("Loading...")
|
| 189 |
+
with gr.Column(scale=2):
|
| 190 |
+
# NEW: Search Engine Selector
|
| 191 |
+
search_engine_selector = gr.Radio(
|
| 192 |
+
list(SEARCH_ENGINES.keys()), value="DuckDuckGo", label="Search Engine"
|
| 193 |
+
)
|
| 194 |
+
with gr.Row():
|
| 195 |
+
new_tab_btn = gr.Button("➕ New Tab"); close_tab_btn = gr.Button("❌ Close Tab")
|
| 196 |
tab_selector = gr.Radio(choices=[], label="Active Tabs", interactive=True)
|
| 197 |
with gr.Accordion("Clickable Links", open=True):
|
| 198 |
links_display = gr.Markdown("...");
|
| 199 |
+
with gr.Row():
|
| 200 |
+
click_num_box = gr.Number(label="Link #", scale=1, minimum=0, step=1)
|
| 201 |
+
click_btn = gr.Button("Click Link", scale=2)
|
| 202 |
|
| 203 |
+
log_display = gr.Textbox(label="Status Log", interactive=False)
|
| 204 |
+
|
| 205 |
all_outputs = [page_content, url_textbox, links_display, tab_selector, log_display]
|
| 206 |
|
| 207 |
+
async def master_handler(current_state, search_engine, action, value=None):
|
| 208 |
global APP_STARTED, P, BROWSER, REVOLVER
|
| 209 |
if not APP_STARTED:
|
| 210 |
print("🚀 First request received, starting up Playwright...");
|
|
|
|
| 212 |
proxy_list_str = os.getenv("PROXY_LIST", ""); REVOLVER = CredentialRevolver(proxy_list_str)
|
| 213 |
print(f"✅ Playwright started. {REVOLVER.count()} proxies loaded."); APP_STARTED = True
|
| 214 |
|
| 215 |
+
new_state, log = await handle_action(current_state, search_engine, action, value)
|
| 216 |
+
ui_updates = update_ui_components(current_state) # Use current state for UI updates
|
| 217 |
|
| 218 |
return (
|
| 219 |
new_state,
|
|
|
|
| 224 |
log
|
| 225 |
)
|
| 226 |
|
| 227 |
+
# Each event listener is its own async function.
|
| 228 |
+
async def on_load(state, search_engine):
|
| 229 |
+
return await master_handler(state, search_engine, "new_tab", None)
|
| 230 |
+
async def on_go_click(state, search_engine, value):
|
| 231 |
+
return await master_handler(state, search_engine, "go", value)
|
| 232 |
+
async def on_click_link(state, search_engine, value):
|
| 233 |
+
return await master_handler(state, search_engine, "click", value)
|
| 234 |
+
async def on_new_tab(state, search_engine):
|
| 235 |
+
return await master_handler(state, search_engine, "new_tab", None)
|
| 236 |
+
async def on_close_tab(state, search_engine):
|
| 237 |
+
return await master_handler(state, search_engine, "close_tab", None)
|
| 238 |
+
async def on_switch_tab(state, search_engine, value):
|
| 239 |
+
return await master_handler(state, search_engine, "switch_tab", value)
|
|
|
|
|
|
|
|
|
|
| 240 |
|
| 241 |
+
# Wire up the event handlers
|
| 242 |
+
go_inputs = [browser_state, search_engine_selector, url_textbox]
|
| 243 |
+
click_inputs = [browser_state, search_engine_selector, click_num_box]
|
| 244 |
+
tab_inputs = [browser_state, search_engine_selector]
|
| 245 |
+
switch_inputs = [browser_state, search_engine_selector, tab_selector]
|
| 246 |
+
outputs = [browser_state, *all_outputs]
|
| 247 |
+
|
| 248 |
+
demo.load(on_load, tab_inputs, outputs)
|
| 249 |
+
go_btn.click(on_go_click, go_inputs, outputs, show_progress="full")
|
| 250 |
+
url_textbox.submit(on_go_click, go_inputs, outputs, show_progress="full")
|
| 251 |
+
click_btn.click(on_click_link, click_inputs, outputs, show_progress="full")
|
| 252 |
+
new_tab_btn.click(on_new_tab, tab_inputs, outputs, show_progress="full")
|
| 253 |
+
close_tab_btn.click(on_close_tab, tab_inputs, outputs)
|
| 254 |
+
tab_selector.input(on_switch_tab, switch_inputs, outputs)
|
| 255 |
|
| 256 |
demo.launch()
|