Workshop-only app: Can an LLM read a PDF and suggest data selections

Ask the Community
1

We have a 100% Workshop application (no Slate or OSDK for this project). Users must open a PDF, read study requirements, then return to the app and choose specific data options based on that PDF.

Question: Is it possible—within Foundry—to use an LLM to (1) read/summarize the PDF and (2) suggest which data points the user should select in the Workshop UI?

Constraints / notes:

  • Workshop only (TypeScript v1), no Slate/OSDK.

  • PDFs may contain sensitive info; we need a secure approach.

  • I’m new to LLMs—any guidance, patterns, or examples are appreciated.

Thanks!

1 Like
2

It is possible by using “AIP generated content“ widget.Where you use output variable as a string.

3

(1) Yes, it is 100% possible to read/summarize PDFs. For an example, I had Foundry read in over 1M PDFs for a Workshop where users search any body text and it returns all PDFs containing the search text or phrase. Additionally, Pipeline Builder (or Code Repository) has an option to use an LLM to get a summary for each PDF.

(2) Yes, Foundry has multiple ways to configure LLM suggestions. Out of curiosity, what is the status of this project? Is Foundry connected to the source of the PDFs? If Foundry can connect to the source of PDFs (i.e. SharePoint or AWS), then I’d recommend a backend solution in Pipeline Builder where an LLM generates a column of confidence scores for the users’ options. If instead the users will be uploading the PDFs, then you might be more successful with something like an AIP Logic function.