Any update ?

samjs•11/5/25, 12:20 AM

Hey @Celestial Rose -- sorry for the slow response. I'm not personally super familiar with this feature, but I can share the feedback with folks internally. You could also try asking over in #ai-search they might have more context.

Celestial RoseOP•11/5/25, 12:22 AM

Hey ! No worries thanks for replying

Celestial RoseOP•11/5/25, 12:23 AM

I would love to circle this internally indeed
There’s currently no way to use pdfs to Markdown for real use-cases because they’re filled with images, and these images don’t get ocr’d at all ! It’s not really AI Search’s fault since the embeddings run on the toMarkdown result

Celestial RoseOP•11/5/25, 12:26 AM

Struggling with this for a while haha https://x.com/rose__celestial/status/1959975514359103501?s=46

Celestial Rose (@Rose__Celestial)

@kristianf_ hey ! quick question, I am using the markdown conversion in the AI binding; just realized that if I pass it a PDF that includes Images, it won't scan the images ... kinda breaking the OCR/Markdown promise. Please tell me there's a backlog task somewhere that will prevent me for

•

8/25/25, 1:45 PM

samjs•11/5/25, 12:28 AM

It’s not really AI Search’s fault since the embeddings run on the toMarkdown result

Ah yeah, but they've been heavily involved with this feature as one of the main users of it

So they might know more about future roadmap for this.

samjs•11/5/25, 12:28 AM

Is the idea that you would want it to effectively call toMarkdowntoMarkdown on the images in the pdf? That would be pretty cool.

Celestial RoseOP•11/5/25, 5:58 PM

Exactly !

Celestial RoseOP•11/5/25, 5:58 PM

Sorry I haven’t seen the reply

Celestial RoseOP•11/5/25, 5:59 PM

I have a lot of documents where there’s critical information on images within the pdfs

Any update ?

Similar Threads

Similar Threads

Similar Threads