I’m excited about serving a shiny app from it!

This is so useful! Thanks for sharing

Oh my!

You choose your own model - I find sonnet very good for R. I have never managed to get the completions to work, just use it in the chat/agentic mode

I find cline to be great

The most complex part was extracting a markdown table from the response so that it could be accessed in other parts of the shiny app (eg download to excel) but this issue showed us how to github.com/posit-dev/sh...

We built a very simple chatbot using shinychat and ellmer and a prompt containing all our quarto blog posts, publication dates and the published url. The large context window of the Gemini models helped a lot! It was very low effort thanks to the great packages but very well received.

Yeah, we do that using quarto. Very useful way to communicate and share with stakeholders at all levels. With ellmer, it was then trivial to turn it into a chatbot which has turned out to be very impactful for identifying gaps in documentation and summarizing team’s accomplishments

Agreed! I have found that Elmer and Shinychat make this really straightforward, so thank you. Joe’s example of a dynamic prompt was really useful for me and enabled a lot of internal apps github.com/jcheng5/r-si...

Searching scientific literature using openalex docs.ropensci.org/openalexR/ar...

Here’s the script! I need to improve it a bit and package it up when I have a bit more time 👉 github.com/victoriano/b...

This taught me that he (and others) weren’t thinking about what problem we were solving (collective sense making of a complex business) and were just following a vendor’s tooling. I lost a lot of respect for a lot of tech folks soon after. This error is pervasive.