Thank you for your question, I don't have experience with Docusaurus but I'll try to list a few ideas, let me know if it can help you.
If you only need raw package metatada (signatures, docs...) that Docusaurus will render somehow at build time you may use my extractor which outputs a plain Javascript object with the collected information (see https://github.com/jsdocs-io/extractor/blob/main/test/packag...).
If you need a pre-rendered page that Docusaurus will collect, you may want to look into https://www.npmjs.com/package/@microsoft/api-extractor which can directly output Markdown files (as well as JSON files) but requires some initial setup to get working.