"Chat with your documents" strategy help needed

[ahasibrm]

I want to create a corpus of academic texts on a subject, usable for research and citation. The citations must include the book title and enough drill-down information to identify the source, eg, "[book name], Section 2, Chapter 3." I think the way to do that is with a sequence of document uploads, each with metadata that includes variables for book, section, chapter.

Questions:

• Can I use one document-loader node and invoke it sequentially, each time manually adjusting the metadata?
• Do I have to use a separate loader node for each doc for a single upsert?
• Is there a way to use the folder upload instead (each doc within the folder having unique metadata values)? How would I assign the metadata?

I've also considered creating a private website wherein each page is its own document, with the page title being the metadata. Then I could use a web scraper node to load the entire site at once. I've never used a web scrape node, so I have questions:

• Is page title associated with the page content such that I could query the model about the source and it would return the page title (metadata)?
• If so, would that result in one concat of metadata rather than the actual three values?
• Is there an alternate way to embed the metadata within the page content?

Lastly, is there a better strategy I haven't thought of?