https://discord.gg/botpress logo
Corrupted Image URLs during Website Crawling & Mar...
# 🤝help
p
Howdy fellow Bot Builders, I want to use a Website-Knowledgebase to display project references in an Autonomous Node. The goal is to have clean image URLs for Card components. **The Issue**: The HTML-to-Markdown conversion during the website crawl occasionally introduces syntax errors. Fragments of Markdown (like )![]() are being appended directly to the indexed image URLs in the Vector Database. Corrupted Format Example:

https://domain.com/image.png▾

)![](https://... Impact: The Autonomous Node retrieves these broken strings. Since the URL is invalid, the images fail to render in the Studio's Card components. Solutions Attempted: I tried to implement a "URL Sanitization" instruction in the Autonomous Node prompt, but it seems to not always work as expected:
Copy code
If an image URL contains trailing characters like ) or !, strictly remove them to ensure the link ends with a valid file extension.
It made it a lot better, but there are still empty image cards sometimes. Thanks in advance ls-sq Screenshots are from the same URL at studio-cloud and the dashboard https://cdn.discordapp.com/attachments/1451158483570524262/1451158484107264010/Screenshot-From-Dashboard_edited.png?ex=697b35a7&is=6979e427&hm=2382746b9e5d5f02b095e02cf5e459f6ff51e00432c52086e317fd47b42c3673& https://cdn.discordapp.com/attachments/1451158483570524262/1451158485067628574/Screenshot-From-Studio_edited.png?ex=697b35a8&is=6979e428&hm=e44e8e0d884d9de0cc895193773400b2dcf64d8f6f7b50f571ae12c9ad55d4c4&
g
write the sanitization in an action and tell the autonomous node to execute it instead of sanitize it.
l
@gentle-policeman-89141 provided a good work-around, can you try it and update us if it works?
p
Yes, it worked, thanks! One minor thing I've noticed: since the latest website auto-recrawl, the Knowledge Base markdown isn't capturing image URLs anymore, it only keeps the alt text. We're fine for now since we don't strictly need the images, but it's a change from how it worked before.
2 Views