Create a wikitext editor user script that adds an OCR button in the Wikitext editor. On clicking the button, the script queries the Google Vision API (via a wrapper API on Tool Labs) with the image of the book on that page (See T140037#2528369 and previous comments there). It populates the content div with the text it got back from the API.
Note that it needs to grab the content language (mw.config.get( 'wgContentLanguage' )) and pass that to the API.
See existing gadget that uses a different OCR service at https://en.wikisource.org/wiki/MediaWiki:Gadget-ocr.js.
Make sure this user script works in both the standard and enhanced edit toolbars. (See code in customiseToolbar() in https://en.wikisource.org/wiki/MediaWiki:Gadget-ocr.js.)