Show HN: Wikimedia and LLM

[jcmp]

Hi! I built this tool that uses the Wikimedia image corpus to enhance LLM-generated text. I originally made it to help explain mathematical concepts during my exam period, but of course, I only finished it once my exams were already over. Anyway, I think there's value in having images alongside text explanations, and since relying solely on AI-generated images isn't so ideal, Wikimedia is a great resource. You can check out an example answer here [1]. If my API key reaches its limit, you'll need your own key to continue using the tool. However, it's completely free, has no ads, requires no sign-up, and if you'd rather not insert your own key, feel free to copy the code from my GitHub or ask Cursor to code it for you.

[yorwba]

Interesting concept!

I see in the debug log that you skip images that exceed 0.5 MB, but some bigger ones actually get through. I guess you increased the limit but didn't update the debug message?

For large images, there are often also resized versions available, it might make sense to use those.

[jcmp]

good catch, actually increased to 1mb, I struggled at the beginning with the 20mb limit for inline data. But will look into the resized versions because i realized sometime is missed great images because of the limit. Thanks for the feedback!

[jcmp]

forgot the url: https://llm.jonathanpagel.com/view?id=8now7ndd5s8ugouxpdjed