[apps/browser] use pyppeteer for stability and speed

apps/browser/app.py

@@ -1,9 +1,5 @@
-from selenium import webdriver
-from selenium.webdriver.chrome.options import Options
-from selenium.webdriver.common.by import By
-from selenium.webdriver.support.ui import WebDriverWait
-from selenium.webdriver.support import expected_conditions as EC
-from selenium.common.exceptions import WebDriverException
+import pyppeteer
+import asyncio
 
 import json
 import hal9 as h9
@@ -14,41 +10,49 @@
 from sitefind import site_find
 from siteuse import site_use
 
-def take_screenshot():
-  time.sleep(2)
-  driver.save_screenshot("screenshot.png")
+async def take_screenshot(page, step):
+  await asyncio.sleep(2)
+  await page.screenshot({'path': "screenshot.png"})
   shutil.copy("screenshot.png", f"storage/screenshot-{int(time.time())}.png")
 
-chrome_options = Options()
-chrome_options.add_argument("--headless")
-chrome_options.add_argument("--no-sandbox")
-chrome_options.add_argument("--disable-dev-shm-usage")
-custom_user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36"
-chrome_options.add_argument(f"user-agent={custom_user_agent}")
+def wrap_in_async_function(code):
+  indented_code = "\n".join("    " + line for line in code.splitlines() if line.strip())  # Indent each line by 4 spaces
+  wrapped_code = f"async def dynamic_async_func(page):\n{indented_code}"
+  return wrapped_code
 
-driver = webdriver.Chrome(options=chrome_options)
+async def main():
+  custom_user_agent = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36"
+  browser = await pyppeteer.launch()
+  page = await browser.newPage()
 
-prompt = h9.input()
-site = site_find(prompt)
+  await page.setUserAgent(custom_user_agent)
 
-print(f"Navigating to {site}")
-driver.get(site)
+  # Get the input and find the site
+  prompt = h9.input()
+  site = site_find(prompt)
 
-for i in range(1, 5):
-  code = "# No code generated"
-  try:
-    code = site_use(prompt, driver.current_url)
-    exec(code)
-    take_screenshot()
-  except WebDriverException as e:
-    print(f"Failed to use browser, driver details follow.\n```\n{code}\n```\n\n```\n{e}\n```\n")
-    print(f"Available Memory: {(psutil.virtual_memory().available/ (1024 ** 2)):.2f} MB")
-  except Exception as e:
-    print(f"Failed to use browser, details follow.\n```\n{code}\n```\n\n```\n{e}\n```\n")
+  print(f"Navigating to {site}")
+  await page.goto(site)
 
-  prompt = h9.input(f"Taking screenshot for step {i}/5, what next?")
+  for i in range(1, 5):
+    code = "# No code generated"
+    try:
+      code = site_use(prompt, page.url)
+      wrapped_code = wrap_in_async_function(code)
+      local_vars = {}
+      exec(wrapped_code, {}, local_vars)
+      await local_vars['dynamic_async_func'](page)
 
-driver.quit()
+      await take_screenshot(page, i)
+    except Exception as e:
+      print(f"Failed to use browser, details follow.\n```\n{code}\n```\n\n```\n{e}\n```\n")
+      print(f"Available Memory: {(psutil.virtual_memory().available/ (1024 ** 2)):.2f} MB")
 
-print("Five tasks completed, this browser session is restarting.")
-print("🌐 I can browse the web, how can I help?")
+    prompt = h9.input(f"Taking screenshot for step {i}/5, what next?")
+
+  await browser.close()
+
+  print("Five tasks completed, this browser session is restarting.")
+  print("🌐 I can browse the web, how can I help?")
+
+asyncio.get_event_loop().run_until_complete(main())

apps/browser/requirements.txt

@@ -1,2 +1,2 @@
-selenium
+pyppeteer
 psutil

apps/browser/siteuse.py

@@ -2,16 +2,14 @@
 from openai import OpenAI
 
 system_prompt = """
-Only write python code using selenium to perform the user request. The code will be run dynamically with eval().
+Only write python code using pyppeteer to perform the user request. The code will be run dynamically with eval().
 
-The driver already stored as a driver variable.
+The page is already stored as a page variable that you can use.
 
-The following includes have been defined:
-from selenium import webdriver
-from selenium.webdriver.chrome.options import Options
-from selenium.webdriver.common.by import By
-from selenium.webdriver.support.ui import WebDriverWait
-from selenium.webdriver.support import expected_conditions as EC
+The following code has already been executed:
+import pyppeteer
+browser = await pyppeteer.launch()
+page = await browser.newPage()
 
 At the beginning of the code, use print() to communicate what the code will do.
 Only reply with a code block for python code.
@@ -20,7 +18,7 @@
 def site_use(prompt, current):
   messages = [
     { "role": "system", "content":  system_prompt},
-    { "role": "user", "content": f"Driver alredy in page {current}. User requests: {prompt}" }
+    { "role": "user", "content": f"Page alredy in page {current}. User requests: {prompt}" }
   ]
   completion = OpenAI().chat.completions.create(model = "gpt-4", messages = messages)
   content = completion.choices[0].message.content