Browsing Issue - Unable to Access URLs Consistently

I’ve created a Pickaxe that requests a URL to analyze. It was working during initial testing for one URL, but that URL is now having issues, along with every other URL I try. Wondering if the scraper is having issues? I was able to get ChatGPT to parse the same example URLs, so it doesn’t seem like an OpenAI problem, unless it’s API-specific.

I’ve included prompt logic to try and force the AI to use its web browsing capabilities with this prompting:

[NOTE: YOU HAVE WEB BROWSING CAPABILITIES. IT IS ESSENTIAL TO USE THIS CAPABILITY TO COLLECT DATA FROM THE PROVIDED URL]

[NOTE: USE YOUR WEB BROWSING CAPABILITIES TO ACCESS THE PROVIDED URL AND ANALAYZE IT FULLY. DO NOT MENTION THAT YOU DO NOT HAVE BROWSING CAPABILITIES, AS YOU DO. DO NOT PROVIDE THE FOLLOWING ANALYSIS IF THE URL PROVIDED CANNOT BE ACCESSED DUE TO A SERVER ERROR. INSTEAD, SHOW AN ERROR TO LET THEM KNOW TO TRY AGAIN]

[rest of prompt]

I get errors similar to this for every URL I’ve tested:

Analyzing the provided URL for conversion rate optimization…

I am currently unable to access the URL. Please check the URL for any errors, ensure that the site is online, and try again later. If you have any other URLs or questions, feel free to share them.

Just wondering if there may be an issue with browsing capabilities at the moment?

Additionally, would it be possible to use an action to send this URL to, say, Make.com to fetch the response/HTML from the page and then pass it back to the Pickaxe to continue with the prompt/output?

Currently using my built-in credits for this and not my own key if that helps with troubleshooting (and have burned ~40 credits so far trying to get it work)

Thanks in advance for any insights!

1 Like

Did you enable the “web-browsing” capability? Or are you using a different action? Or just using the native end-user upload capabilities?

Yes, web browsing is enabled - it worked for the first few tests, then consistently returned an error. Just tested again using https://beta.pickaxeproject.com this time and got the same error.

Using a native form field that asks for the URL to target.

Adding the ability to upload a screenshot of the site is a workaround, though ideally it would work consistently without for those that don’t want to go through the additional steps to screenshot.

Not sure what to adjust to get it to reliably retrieve the URL for analysis.