Allow for multiple output

[CChen89]

Hi,

Thank you for developing these excellent packages. As a beginner with ellmer and shinychat, I'm interested in learning how to return multiple outputs in a single chat round via tool calling.

In R/Shiny, returning a list (e.g., return(list(A = fig1, B = fig2, C = data.frame(), D = Text))) is the typical approach. Is it possible for ellmer's Content() function to bundle multiple output types in one response, like:

Content(
  ContentImage(fig1),
  ContentImage(fig2),
  ContentTable(my_table),
  ContentText(glue::glue('My summary text here'))
)

In my use case, my tool calling function needs to output and shinychatbox should display:

1. Two figures (ggplot objects)
2. One table (e.g., a gt HTML table)
3. One character sentence (e.g., using glue with variables)

I haven't been able to find documentation or tutorials covering this multi-output approach in ellmer/shinychat. I did read your post at #735, but that covers only figures.

I also saw a shinychat article covering ContentToolResults() (https://posit-dev.github.io/shinychat/r/articles/tool-ui.html#alternative-display-content), but it seems to only allow a sentence and table via an extra argument, rather than arbitrary outputs.

Please let me know if this question would be better suited for Stack Overflow.

Thank you for any guidance!

[hadley]

In general, #735 is your go to — it's not really that ellmer doesn't support this, it's that most LLMs do not support this.