Update app.py

app.py

@@ -1,5 +1,6 @@
 import os
 os.system("playwright install")
+import os
 import re
 import urllib.parse
 import asyncio
@@ -11,34 +12,25 @@ from playwright.async_api import async_playwright
 
 # --- 1. GLOBAL RESOURCES & CONFIGURATION ---
 
-# This dictionary will hold the long-lived Playwright and Browser objects
+# This dictionary will hold the long-lived Playwright and Browser objects.
+# It starts empty and browsers are added on-demand.
 PLAYWRIGHT_STATE: Dict = {}
 
-# EXPANDED: A comprehensive list of search engines
+# A comprehensive list of search engines
 SEARCH_ENGINES = {
-    # Original
-    "DuckDuckGo": "https://duckduckgo.com/html/?q={query}",
-    "Google": "https://www.google.com/search?q={query}",
-    "Bing": "https://www.bing.com/search?q={query}",
-    "Brave": "https://search.brave.com/search?q={query}",
-    "Ecosia": "https://www.ecosia.org/search?q={query}",
-    # 10 More Added
-    "Yahoo": "https://search.yahoo.com/search?p={query}",
-    "Startpage": "https://www.startpage.com/sp/search?q={query}",
-    "Qwant": "https://www.qwant.com/?q={query}",
-    "Swisscows": "https://swisscows.com/web?query={query}",
-    "You.com": "https://you.com/search?q={query}",
-    "SearXNG": "https://searx.be/search?q={query}",
-    "MetaGer": "https://metager.org/meta/meta.ger-en?eingabe={query}",
-    "Yandex": "https://yandex.com/search/?text={query}",
-    "Baidu": "https://www.baidu.com/s?wd={query}",
+    "DuckDuckGo": "https://duckduckgo.com/html/?q={query}", "Google": "https://www.google.com/search?q={query}",
+    "Bing": "https://www.bing.com/search?q={query}", "Brave": "https://search.brave.com/search?q={query}",
+    "Ecosia": "https://www.ecosia.org/search?q={query}", "Yahoo": "https://search.yahoo.com/search?p={query}",
+    "Startpage": "https://www.startpage.com/sp/search?q={query}", "Qwant": "https://www.qwant.com/?q={query}",
+    "Swisscows": "https://swisscows.com/web?query={query}", "You.com": "https://you.com/search?q={query}",
+    "SearXNG": "https://searx.be/search?q={query}", "MetaGer": "https://metager.org/meta/meta.ger-en?eingabe={query}",
+    "Yandex": "https://yandex.com/search/?text={query}", "Baidu": "https://www.baidu.com/s?wd={query}",
     "Perplexity": "https://www.perplexity.ai/search?q={query}"
 }
 
-
 # --- 2. ADVANCED HTML-TO-MARKDOWN CONVERTER (Unchanged) ---
 class HTML_TO_MARKDOWN_CONVERTER:
-    # ... [The class code is identical to the previous version and remains unchanged] ...
+    # ... [The class code is identical and correct] ...
     def __init__(self, soup: BeautifulSoup, base_url: str):
         self.soup = soup
         self.base_url = base_url
@@ -97,65 +89,69 @@ class HTML_TO_MARKDOWN_CONVERTER:
         return inner_md
 
 
-# --- 3. CORE API FUNCTION ---
-
-async def initialize_playwright():
-    """Launches Playwright and browser instances if they don't already exist."""
-    if "playwright" not in PLAYWRIGHT_STATE:
-        print("🚀 First request received, starting up Playwright...")
-        p = await async_playwright().start()
-        PLAYWRIGHT_STATE["playwright"] = p
-        PLAYWRIGHT_STATE["chromium"] = await p.chromium.launch(headless=True)
-        PLAYWRIGHT_STATE["firefox"] = await p.firefox.launch(headless=True)
-        PLAYWRIGHT_STATE["webkit"] = await p.webkit.launch(headless=True)
-        print("✅ Playwright and browsers are ready.")
+# --- 3. CORE API FUNCTION (WITH LAZY LOADING) ---
 
 async def perform_web_browse(query: str, browser_name: str, search_engine: str):
     """
     A stateless function that takes a query, browser, and search engine,
     then returns the parsed content of the resulting page.
+    It launches and caches browsers on-demand.
     """
-    await initialize_playwright()
+    # Step 1: Initialize Playwright process itself if not already running.
+    if "playwright" not in PLAYWRIGHT_STATE:
+        print("🚀 First request received, starting Playwright process...")
+        PLAYWRIGHT_STATE["playwright"] = await async_playwright().start()
+        print("✅ Playwright process is running.")
+
+    # Step 2: Check if the *specific browser requested* has been launched.
+    browser_key = browser_name.lower()
+    if browser_key not in PLAYWRIGHT_STATE:
+        print(f"🚀 Launching '{browser_key}' for the first time...")
+        try:
+            p = PLAYWRIGHT_STATE["playwright"]
+            if browser_key == 'firefox':
+                browser_instance = await p.firefox.launch(headless=True)
+            elif browser_key == 'chromium':
+                browser_instance = await p.chromium.launch(headless=True)
+            elif browser_key == 'webkit':
+                browser_instance = await p.webkit.launch(headless=True)
+            else:
+                raise ValueError(f"Invalid browser name: {browser_name}")
+            PLAYWRIGHT_STATE[browser_key] = browser_instance
+            print(f"✅ '{browser_key}' is now running and cached.")
+        except Exception as e:
+            error_message = str(e).splitlines()[0]
+            print(f"❌ Failed to launch '{browser_key}': {error_message}")
+            return {"status": "error", "query": query, "error_message": f"Failed to launch browser '{browser_key}'. Your system might be missing dependencies. Error: {error_message}"}
+    
+    browser_instance = PLAYWRIGHT_STATE[browser_key]
 
-    # Determine if the query is a URL or a search term
+    # Step 3: Determine URL
     is_url = urllib.parse.urlparse(query).scheme in ['http', 'https']
     if is_url:
         url = query
     else:
         search_url_template = SEARCH_ENGINES.get(search_engine)
         if not search_url_template:
-            return {"error": f"Invalid search engine: '{search_engine}'. Please choose from the provided list."}
+            return {"error": f"Invalid search engine: '{search_engine}'."}
         url = search_url_template.format(query=urllib.parse.quote_plus(query))
 
-    browser_instance = PLAYWRIGHT_STATE.get(browser_name.lower())
-    if not browser_instance:
-        return {"error": f"Invalid browser: '{browser_name}'. Use 'chromium', 'firefox', or 'webkit'."}
-
+    # Step 4: Create isolated context and browse
     context = await browser_instance.new_context(user_agent='Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36')
     page = await context.new_page()
 
     try:
         print(f"Navigating to: {url} using {browser_name}...")
         await page.goto(url, wait_until='domcontentloaded', timeout=30000)
-        
-        final_url = page.url
-        title = await page.title() or "No Title"
+        final_url, title = page.url, await page.title() or "No Title"
         print(f"Arrived at: {final_url}")
         
-        html_content = await page.content()
-        soup = BeautifulSoup(html_content, 'lxml')
-        
+        soup = BeautifulSoup(await page.content(), 'lxml')
         converter = HTML_TO_MARKDOWN_CONVERTER(soup, base_url=final_url)
         markdown_text = converter.convert()
         
         print("Content parsed successfully.")
-        return {
-            "status": "success",
-            "query": query,
-            "final_url": final_url,
-            "page_title": title,
-            "markdown_content": markdown_text,
-        }
+        return {"status": "success", "query": query, "final_url": final_url, "page_title": title, "markdown_content": markdown_text}
     except Exception as e:
         error_message = str(e).splitlines()[0]
         print(f"An error occurred: {error_message}")
@@ -166,42 +162,18 @@ async def perform_web_browse(query: str, browser_name: str, search_engine: str):
         print("Session context closed.")
 
 
-# --- 4. GRADIO INTERFACE & API LAUNCH ---
-
+# --- 4. GRADIO INTERFACE & API LAUNCH (Unchanged) ---
 with gr.Blocks(title="Web Browse API", theme=gr.themes.Soft()) as demo:
+    # ... UI definition is identical ...
     gr.Markdown("# Web Browse API")
-    gr.Markdown(
-        "This interface exposes a stateless API endpoint (`/api/web_browse`) to fetch and parse web content."
-    )
-    
-    query_input = gr.Textbox(
-        label="URL or Search Query", 
-        placeholder="e.g., https://openai.com or 'history of artificial intelligence'"
-    )
-    
+    gr.Markdown("This interface exposes a stateless API endpoint (`/api/web_browse`) to fetch and parse web content.")
+    query_input = gr.Textbox(label="URL or Search Query", placeholder="e.g., https://openai.com or 'history of artificial intelligence'")
     with gr.Row():
-        browser_input = gr.Dropdown(
-            label="Browser", 
-            choices=["firefox", "chromium", "webkit"], 
-            value="firefox",
-            scale=1
-        )
-        search_engine_input = gr.Dropdown(
-            label="Search Engine (for non-URL queries)",
-            choices=sorted(list(SEARCH_ENGINES.keys())),
-            value="DuckDuckGo",
-            scale=2
-        )
-    
+        browser_input = gr.Dropdown(label="Browser", choices=["firefox", "chromium", "webkit"], value="firefox", scale=1)
+        search_engine_input = gr.Dropdown(label="Search Engine (for non-URL queries)", choices=sorted(list(SEARCH_ENGINES.keys())), value="DuckDuckGo", scale=2)
     submit_button = gr.Button("Browse", variant="primary")
     output_json = gr.JSON(label="API Result")
-    
-    submit_button.click(
-        fn=perform_web_browse,
-        inputs=[query_input, browser_input, search_engine_input],
-        outputs=output_json,
-        api_name="web_browse" # Creates the POST /api/web_browse endpoint
-    )
+    submit_button.click(fn=perform_web_browse, inputs=[query_input, browser_input, search_engine_input], outputs=output_json, api_name="web_browse")
 
 if __name__ == "__main__":
     demo.launch()