Hi all! I've been building a small R package and would love your feedback.
You give it your own documents (PDFs, Word files, web pages) and ask questions in a Shiny app. It answers from what it read, and says so when the documents don't cover something. You can also train a model on the same documents, all on your own machine.
On Costco's 76-page annual report it got 12 of 13 questions right.
Write-up: Fine-tune your own model from R, starting from your documents • dragonfarm
What would you try it on? And if you do try it, what felt confusing?
