Make web browser agent example a CLI script #416
Conversation
|
@aymeric-roucher I wonder if it's helium or Qwen2VL but I can't seem to even execute |
|
fixes #416 |
|
@aymeric-roucher is there any other option you'd like me to add to argparse? |
| @@ -64,7 +86,7 @@ def save_screenshot(step_log: ActionStep, agent: CodeAgent) -> None: | |||
| # Initialize driver and agent | |||
| chrome_options = webdriver.ChromeOptions() | |||
| chrome_options.add_argument("--force-device-scale-factor=1") | |||
| chrome_options.add_argument("--window-size=1000,1300") | |||
| chrome_options.add_argument("--window-size=1000,1350") | |||
There was a problem hiding this comment.
Why this change ? (out of curiosity)
| parser.add_argument( | ||
| "--model-id", | ||
| type=str, | ||
| default="gpt-4o", |
There was a problem hiding this comment.
Made GPT-4o default here while waiting for Qwen-2.5-VL!
| print(f"Error handling selector {selector}: {str(e)}") | ||
| continue | ||
| return "Modals closed" | ||
| webdriver.ActionChains(driver).send_keys(Keys.ESCAPE).perform() |
There was a problem hiding this comment.
Vastly simplified this tool thanks to this comment on our blog post: just using Escape works without heavy engineering!
|
|
||
| agent.run(search_request + helium_instructions) | ||
| agent.python_executor("from helium import *", agent.state) |
There was a problem hiding this comment.
Found this trick to avoid any potential errors: forces running the imports at start!
There was a problem hiding this comment.
this is very cool! thank you!
aymeric-roucher
changed the title
aymeric-roucher
changed the title
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
This PR makes computer use example into a CLI script for more general easy use + adds
HfApiModelas provider since we'll have longer context Qwen2VL soon