Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
import os
|
2 |
os.system("playwright install")
|
3 |
-
# app.py (Final Version with
|
4 |
|
5 |
import gradio as gr
|
6 |
from playwright.async_api import async_playwright
|
@@ -17,7 +17,6 @@ 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}",
|
@@ -41,92 +40,49 @@ class CredentialRevolver:
|
|
41 |
self.proxies = self._parse_proxies(proxy_string)
|
42 |
self.proxy_cycler = cycle(self.proxies) if self.proxies else None
|
43 |
def _parse_proxies(self, proxy_string: str):
|
44 |
-
proxies = []
|
45 |
for line in proxy_string.strip().splitlines():
|
46 |
try:
|
47 |
-
parsed = urllib.parse.urlparse(f"//{line.strip()}")
|
48 |
-
server = f"{parsed.scheme or 'http'}://{parsed.hostname}:{parsed.port}"
|
49 |
proxies.append({"server": server, "username": parsed.username, "password": parsed.password})
|
50 |
except: pass
|
51 |
return proxies
|
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
|
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 |
-
|
78 |
-
if
|
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 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
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:
|
@@ -136,19 +92,11 @@ async def _fetch_and_update_tab_state(tab_state: TabState, url: str):
|
|
136 |
return log
|
137 |
|
138 |
async def handle_action(browser_state: BrowserState, search_engine: str, action: str, value=None):
|
139 |
-
|
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)
|
@@ -156,7 +104,7 @@ async def handle_action(browser_state: BrowserState, search_engine: str, action:
|
|
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/")
|
160 |
elif action == "click" and active_tab_state and value is not None:
|
161 |
try:
|
162 |
link_index = int(value)
|
@@ -172,36 +120,39 @@ async def handle_action(browser_state: BrowserState, search_engine: str, action:
|
|
172 |
else: log = "Cannot close the last tab."
|
173 |
elif action == "switch_tab" and value is not None:
|
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 |
-
|
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 |
-
|
191 |
-
|
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):
|
@@ -213,44 +164,24 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
|
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(
|
217 |
|
218 |
-
return (
|
219 |
-
new_state,
|
220 |
-
ui_updates[page_content],
|
221 |
-
ui_updates[url_textbox],
|
222 |
-
ui_updates[links_display],
|
223 |
-
ui_updates[tab_selector],
|
224 |
-
log
|
225 |
-
)
|
226 |
|
227 |
-
|
228 |
-
async def
|
229 |
-
|
230 |
-
async def
|
231 |
-
|
232 |
-
async def
|
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 |
-
|
249 |
-
|
250 |
-
|
251 |
-
|
252 |
-
|
253 |
-
|
254 |
-
tab_selector.input(on_switch_tab, switch_inputs, outputs)
|
255 |
|
256 |
demo.launch()
|
|
|
1 |
import os
|
2 |
os.system("playwright install")
|
3 |
+
# app.py (Final, Working Async Version with All Bugs Fixed)
|
4 |
|
5 |
import gradio as gr
|
6 |
from playwright.async_api import async_playwright
|
|
|
17 |
LIVE_CONTEXTS = {}
|
18 |
APP_STARTED = False
|
19 |
|
|
|
20 |
SEARCH_ENGINES = {
|
21 |
"DuckDuckGo": "https://duckduckgo.com/html/?q={query}",
|
22 |
"Google": "https://www.google.com/search?q={query}",
|
|
|
40 |
self.proxies = self._parse_proxies(proxy_string)
|
41 |
self.proxy_cycler = cycle(self.proxies) if self.proxies else None
|
42 |
def _parse_proxies(self, proxy_string: str):
|
43 |
+
proxies = [];
|
44 |
for line in proxy_string.strip().splitlines():
|
45 |
try:
|
46 |
+
parsed = urllib.parse.urlparse(f"//{line.strip()}"); server = f"{parsed.scheme or 'http'}://{parsed.hostname}:{parsed.port}"
|
|
|
47 |
proxies.append({"server": server, "username": parsed.username, "password": parsed.password})
|
48 |
except: pass
|
49 |
return proxies
|
50 |
def get_next(self): return next(self.proxy_cycler) if self.proxy_cycler else None
|
51 |
def count(self): return len(self.proxies) if self.proxies else 0
|
52 |
|
53 |
+
# --- 3. CORE ASYNC LOGIC & FORMATTING ---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
def _process_element_to_markdown(element):
|
55 |
+
if isinstance(element, NavigableString): return element.strip()
|
56 |
+
if element.name is None: return ''
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
inner_text = ''.join(_process_element_to_markdown(child) for child in element.children)
|
58 |
+
if element.name in ['p', 'div', 'article', 'section']: return f"\n{inner_text.strip()}\n"
|
59 |
+
if element.name == 'h1': return f"\n# {inner_text.strip()}\n"
|
60 |
+
if element.name == 'h2': return f"\n## {inner_text.strip()}\n"
|
61 |
+
if element.name == 'h3': return f"\n### {inner_text.strip()}\n"
|
62 |
+
if element.name in ['h4', 'h5', 'h6']: return f"\n#### {inner_text.strip()}\n"
|
63 |
+
if element.name == 'li': return f"* {inner_text.strip()}\n"
|
64 |
+
if element.name in ['ul', 'ol']: return f"\n{inner_text}\n"
|
65 |
+
if element.name in ['strong', 'b']: return f"**{inner_text.strip()}**"
|
66 |
+
if element.name in ['em', 'i']: return f"*{inner_text.strip()}*"
|
67 |
+
if element.name in ['pre', 'code']: return f"\n```\n{inner_text.strip()}\n```\n"
|
68 |
+
if element.name == 'a': return f"[{inner_text.strip()}]({element.get('href', '')})"
|
69 |
+
if element.name == 'hr': return "\n---\n"
|
70 |
+
return inner_text
|
71 |
|
72 |
+
def _format_html_to_markdown(soup):
|
73 |
+
content_node = soup.find('main') or soup.find('body')
|
74 |
+
if not content_node: return "Could not find main body content."
|
75 |
+
for el in content_node.select('nav, footer, header, aside, form, script, style'): el.decompose()
|
76 |
+
return _process_element_to_markdown(content_node)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
77 |
|
78 |
async def _fetch_and_update_tab_state(tab_state: TabState, url: str):
|
|
|
79 |
log = f"▶️ Navigating to {url}..."; live_page = LIVE_CONTEXTS[tab_state.id]["page"]
|
80 |
try:
|
81 |
await live_page.goto(url, wait_until='domcontentloaded', timeout=30000)
|
82 |
tab_state.url = live_page.url; tab_state.title = await live_page.title() or "No Title"
|
83 |
log += f"\n✅ Arrived at: {tab_state.url}"
|
84 |
+
html_content = await live_page.content(); soup = BeautifulSoup(html_content, 'lxml')
|
|
|
|
|
|
|
|
|
85 |
tab_state.parsed_text = _format_html_to_markdown(soup)
|
|
|
86 |
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')]
|
87 |
log += f"\n🔗 Found {len(tab_state.links)} links."
|
88 |
except Exception as e:
|
|
|
92 |
return log
|
93 |
|
94 |
async def handle_action(browser_state: BrowserState, search_engine: str, action: str, value=None):
|
95 |
+
active_tab_state = browser_state.get_active_tab()
|
|
|
|
|
96 |
if action == "go" and active_tab_state and value:
|
97 |
is_url = urllib.parse.urlparse(value).scheme and urllib.parse.urlparse(value).netloc
|
98 |
+
url = value if is_url else SEARCH_ENGINES.get(search_engine, SEARCH_ENGINES["DuckDuckGo"]).format(query=urllib.parse.quote_plus(value))
|
|
|
|
|
|
|
|
|
99 |
log = await _fetch_and_update_tab_state(active_tab_state, url)
|
|
|
|
|
100 |
elif action == "new_tab":
|
101 |
tab_id, proxy_config = str(uuid.uuid4()), REVOLVER.get_next()
|
102 |
context = await BROWSER.new_context(proxy=proxy_config)
|
|
|
104 |
LIVE_CONTEXTS[tab_id] = {"context": context, "page": page}
|
105 |
new_tab = TabState(tab_id, proxy_used=proxy_config['server'].split('@')[-1] if proxy_config else "Direct")
|
106 |
browser_state.tabs.append(new_tab); browser_state.active_tab_id = tab_id
|
107 |
+
log = await _fetch_and_update_tab_state(new_tab, "https://www.startpage.com/")
|
108 |
elif action == "click" and active_tab_state and value is not None:
|
109 |
try:
|
110 |
link_index = int(value)
|
|
|
120 |
else: log = "Cannot close the last tab."
|
121 |
elif action == "switch_tab" and value is not None:
|
122 |
browser_state.active_tab_id = value; log = f"Switched to tab."
|
123 |
+
else: log = "No action taken."
|
124 |
return browser_state, log
|
125 |
|
126 |
+
# ** CRITICAL BUG FIX: `NameError` is fixed by defining this function before it is called **
|
127 |
+
def update_ui_components(browser_state: BrowserState):
|
128 |
+
active_tab = browser_state.get_active_tab()
|
129 |
+
if not active_tab: return {page_content: gr.Markdown("No active tabs."), url_textbox: "", links_display: "", tab_selector: gr.Radio(choices=[])}
|
130 |
+
tab_choices = [(f"Tab {i}: {t.title[:25]}... ({t.proxy_used})", t.id) for i, t in enumerate(browser_state.tabs)]
|
131 |
+
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._")
|
132 |
+
page_md = f"# {active_tab.title}\n**URL:** {active_tab.url}\n\n---\n\n{active_tab.parsed_text}"
|
133 |
+
return {
|
134 |
+
page_content: gr.Markdown(page_md),
|
135 |
+
url_textbox: gr.Textbox(value=active_tab.url), links_display: gr.Markdown(links_md),
|
136 |
+
tab_selector: gr.Radio(choices=tab_choices, value=active_tab.id, label="Active Tabs"),
|
137 |
+
}
|
138 |
+
|
139 |
# --- 4. GRADIO UI AND EVENT HANDLING ---
|
140 |
with gr.Blocks(theme=gr.themes.Soft(), title="Real Browser Demo") as demo:
|
141 |
browser_state = gr.State(BrowserState())
|
142 |
+
gr.Markdown("# 🛰️ Real Browser Demo v2.1")
|
|
|
143 |
with gr.Row():
|
144 |
with gr.Column(scale=4):
|
145 |
+
with gr.Row(): url_textbox = gr.Textbox(label="Enter URL or Search Query", interactive=True, scale=4); go_btn = gr.Button("Go", variant="primary", scale=1)
|
146 |
+
with gr.Accordion("Page Content (Formatted)", open=True): page_content = gr.Markdown("Loading...")
|
|
|
|
|
|
|
147 |
with gr.Column(scale=2):
|
148 |
+
search_engine_selector = gr.Radio(list(SEARCH_ENGINES.keys()), value="DuckDuckGo", label="Search Engine")
|
149 |
+
with gr.Row(): new_tab_btn = gr.Button("➕ New Tab"); close_tab_btn = gr.Button("❌ Close Tab")
|
|
|
|
|
|
|
|
|
150 |
tab_selector = gr.Radio(choices=[], label="Active Tabs", interactive=True)
|
151 |
with gr.Accordion("Clickable Links", open=True):
|
152 |
links_display = gr.Markdown("...");
|
153 |
+
with gr.Row(): click_num_box = gr.Number(label="Link #", scale=1, minimum=0, step=1); click_btn = gr.Button("Click Link", scale=2)
|
|
|
|
|
|
|
154 |
log_display = gr.Textbox(label="Status Log", interactive=False)
|
155 |
+
|
156 |
all_outputs = [page_content, url_textbox, links_display, tab_selector, log_display]
|
157 |
|
158 |
async def master_handler(current_state, search_engine, action, value=None):
|
|
|
164 |
print(f"✅ Playwright started. {REVOLVER.count()} proxies loaded."); APP_STARTED = True
|
165 |
|
166 |
new_state, log = await handle_action(current_state, search_engine, action, value)
|
167 |
+
ui_updates = update_ui_components(new_state)
|
168 |
|
169 |
+
return (new_state, ui_updates[page_content], ui_updates[url_textbox], ui_updates[links_display], ui_updates[tab_selector], log)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
170 |
|
171 |
+
async def on_load(state, search_engine): return await master_handler(state, search_engine, "new_tab", None)
|
172 |
+
async def on_go_click(state, search_engine, value): return await master_handler(state, search_engine, "go", value)
|
173 |
+
async def on_click_link(state, search_engine, value): return await master_handler(state, search_engine, "click", value)
|
174 |
+
async def on_new_tab(state, search_engine): return await master_handler(state, search_engine, "new_tab", None)
|
175 |
+
async def on_close_tab(state, search_engine): return await master_handler(state, search_engine, "close_tab", None)
|
176 |
+
async def on_switch_tab(state, search_engine, value): return await master_handler(state, search_engine, "switch_tab", value)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
|
|
|
|
|
|
|
|
|
|
|
178 |
outputs = [browser_state, *all_outputs]
|
179 |
+
demo.load(on_load, [browser_state, search_engine_selector], outputs)
|
180 |
+
go_btn.click(on_go_click, [browser_state, search_engine_selector, url_textbox], outputs, show_progress="full")
|
181 |
+
url_textbox.submit(on_go_click, [browser_state, search_engine_selector, url_textbox], outputs, show_progress="full")
|
182 |
+
click_btn.click(on_click_link, [browser_state, search_engine_selector, click_num_box], outputs, show_progress="full")
|
183 |
+
new_tab_btn.click(on_new_tab, [browser_state, search_engine_selector], outputs, show_progress="full")
|
184 |
+
close_tab_btn.click(on_close_tab, [browser_state, search_engine_selector], outputs)
|
185 |
+
tab_selector.input(on_switch_tab, [browser_state, search_engine_selector, tab_selector], outputs)
|
|
|
186 |
|
187 |
demo.launch()
|