This is a demonstration of using an LLM to enhance a data dashboard written in Shiny.
Live demo (Python version)
To run locally, you'll need to create an .env file in the repo root with OPENAI_API_KEY= followed by a valid OpenAI API key. Or if that environment value is set some other way, you can skip the .env file.
Then run:
pak::pak(c("base64enc", "bslib", "DBI", "dplyr", "duckdb", "fastmap",
"fontawesome", "ggplot2", "ggridges", "here", "mirai", "irudnyts/openai@r6",
"plotly", "promises", "reactable", "shiny"))This app sends at least your data schema to a remote LLM. As written, it also permits the LLM to run SQL queries against your data and get the results back. Please keep these facts in mind when dealing with sensitive data.
This app currently has a slightly more limited chat experience than the equivalent Python version (see below). I hope to bring these experiences to parity as the LLM client packages in R mature
You can find the Python version of this app (including live demo) at https://github.com/jcheng5/py-sidebot.