Import guides
The real path is a folder tree of individually exported pages, one file per page, keeping every page's place in your notebook's own hierarchy. OneNote has no menu item for this itself: producing it means running a script against the same GetHierarchy and Publish automation calls OneNote's desktop app exposes, for example meichthys/onenote-html-export, a free, open-source tool built for exactly this. It needs the desktop app on Windows; there is no web or Mac equivalent for producing the tree. Point Canonry at the folder it produces, or a zip of it.
If you cannot produce that tree, export the whole notebook or a section to PDF or DOCX from the Windows desktop app instead, or to .onepkg from OneNote on the web with a personal Microsoft account, and hand Canonry that file. It is read through the PDF or DOCX guide, the same as any other PDF or DOCX, keeping every page but losing the notebook's own hierarchy: a subpage becomes just another heading once the export flattens it.
Canonry looks for a tree of HTML pages where at least one page has its own sibling folder of embedded attachments, named after the page with _files appended - the shape only an exported OneNote page tree produces, and nothing else mimics. Find that and it shows you how many pages it counted before asking you to confirm the OneNote playbook; say no and it falls back to a short list of other playbooks. A notebook where no page embeds an image has no such folder for Canonry to key on, so detection will not recognise it as OneNote - bring in at least one page with an embedded image if you can, or use the PDF, DOCX or .onepkg fallback above.
The notebook's own hierarchy is what a flattened PDF or DOCX export throws away, and it is the reason this path exists: a page sitting in a folder named after another page is proposed as a subpage of it, with the folder tree itself standing as the evidence, since OneNote's own export produced it rather than anything written in the page itself. Every link to another page in the tree becomes a candidate relation the same way, and every embedded image travels across as an attachment on the page's own entity. A page's title becomes its entity name.