Importing/scraping recipe failed, now what?

[InfiniteCactusDev]

New to Mealie. One of the first recipes I tried importing, and it failed hard. Besides a single line of text it got nothing. Here's the recipe: https://www.thefrenchcookingacademy.com/recipes/basque-cheesecake
It seems like such a simple and clean page which should be easy to import.
I tried looking for documentation on the parser used and how to extend it, but was unable to find this.
So my question basically is - if I want to improve the parser where should I look? Is there a place to request sites to be supported?

[michael-genson]

Under the hood Mealie uses the recipe-scrapers library. If it's not supported there we probably won't get much here (there are exceptions to this). In general the reason we can/can't scrape a website boils down to:

1. Does it have the right metadata?
2. Is it supported by recipe-scrapers?

If you look at the source of the website you'll find it has no recipe metadata (i.e. https://schema.org/Recipe) so there's no way to scrape it normally. However we do support OpenAI as a fallback in these scenarios, and I'm guessing it would have no problem with this site. More info in the docs on how to set that up.

[InfiniteCactusDev]

Thanks for the in depth response!

It makes sense (now) that the website isn't supported by the scraper library.

On a quest to try and getting this specific website to work I've now installed Ollama with model qwen3:8b. The tests on the Debug page work correctly. It also parsed an ingredient line successfully.

However, trying to import the recipe either via url or by manually coping the html does not work at all. Trying to debug the scraper (with openai enabled) gives "recipe_scrapers was unable to scrape this URL".

What would the next steps be?

[michael-genson]

We've had users report limited success with self hosted models, typically boiling down to context windows not being large enough and hardware not being beefy enough. You'll have to check out some other Ollama threads for some ideas on next steps.

If you decide to switch to OpenAI, the costs are quite minimal, scraping 10 websites with OpenAI is something like 10-20 cents using GPT-5 (and most sites won't require OpenAI).