Typesense operates on JSON documents. So you'd have to first extract the data from PDFs / office docs into JSON and then index it.

There are libraries that let you do the conversion:

https://github.com/modesty/pdf2json

https://github.com/microsoft/Simplify-Docx