- Issue created by @baikho
- 🇫🇮Finland merilainen
I think this is a feature request, because it's possible to use an intermediate field called something like "Source HTML with absolute img urls" where AI can be instructed to change any relative URLs to absolute with the following prompt:
I will give you a page HTML source as the context and I want you to find all relative img tags, replace them in the source with prefix https://boonstoppelverf.nl/ domain so that the url will become absolute. Return the rest of the source HTML as is. If the context is empty, do not do anything.
Context: {{ raw_context }}
It's not ideal and wastes tokens, but it works.
- 🇫🇮Finland merilainen
Here is a patch which adds support for relative asset paths for both File and Image extractors.
This is implemented by providing a "domain" configuration which can be set when relative paths should be processed. Leaving it empty will skip processing of relative paths.
Other changes:
- Dynamic extensions support for ImageExctractor (copied from FileExtractor)
- Use extraFormFields instead of extraAdvancedFormFields in both extractors (makes configurable parts more visible)