Merge pull request #126 from posit-dev/feat/evaluate-tool

feat: A tool to run R code

Author: gadenbuie

.Rbuildignore

@@ -16,3 +16,5 @@
 ^scripts$
 ^\.claude$
 ^CRAN-SUBMISSION$
+^\.prettierrc$
+^node_modules$

.github/workflows/R-CMD-check.yaml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ${{ matrix.config.os }}
 
     name: ${{ matrix.config.os }} (${{ matrix.config.r }})
-    timeout-minutes: 45
+    timeout-minutes: 180
 
     strategy:
       fail-fast: false

.prettierrc

@@ -0,0 +1,14 @@
+{
+  "organizeImportsSkipDestructiveCodeActions": true,
+  "singleQuote": false,
+  "semi": false,
+  "trailingComma": "all",
+  "overrides": [
+    {
+      "files": "**/*.scss",
+      "options": {
+        "printWidth": 150
+      }
+    }
+  ]
+}

DESCRIPTION

@@ -53,10 +53,13 @@ Suggests:
     chromote,
     DBI,
     duckdb,
+    evaluate,
+    fansi,
     gert,
     gh,
     htmltools,
     pandoc,
+    ragg,
     shiny,
     shinychat (>= 0.2.0),
     testthat (>= 3.0.0),
@@ -93,6 +96,7 @@ Collate:
     'tool-git.R'
     'tool-github.R'
     'tool-rstudioapi.R'
+    'tool-run.R'
     'tool-search-packages.R'
     'tool-session-package-installed.R'
     'tool-sessioninfo.R'

NAMESPACE

@@ -47,6 +47,7 @@ export(btw_tool_git_log)
 export(btw_tool_git_status)
 export(btw_tool_github)
 export(btw_tool_ide_read_current_editor)
+export(btw_tool_run_r)
 export(btw_tool_search_package_info)
 export(btw_tool_search_packages)
 export(btw_tool_session_check_package_installed)

NEWS.md

@@ -1,5 +1,7 @@
 # btw (development version)
 
+* New `btw_tool_run_r()` tool allows LLMs to run R code and to see the output, including of plots. Because this tool lets LLMs run R arbitrary R code in the global environment (which can be great but can also have security implications), it is opt-in and disabled by default. See `?btw_tool_run_r` for more details (#126).
+
 * `btw_tool_docs_help_page()` now uses markdown headings and sections for argument descriptions, rather than a table. This is considerably more token efficient when the argument descriptions have more than one paragraph and can't be converted into a markdown table (@jeanchristophe13v, #123).
 
 * btw now removes large inline base64-encoded images, replacing them with a placeholder containing the image's alt text (@jeanchristophe13v, #119).

R/btw_client.R

@@ -133,7 +133,7 @@ btw_client <- function(
 }
 
 btw_client_config <- function(client = NULL, tools = NULL, config = list()) {
-  config$options <- flatten_config_options(config$options)
+  # Options should be flattened and btw-prefixed by `read_btw_file()`.
   withr::local_options(config$options)
 
   config$tools <-
@@ -275,7 +275,11 @@ flatten_config_options <- function(opts, prefix = "btw", sep = ".") {
       }
 
       for (i in seq_along(x)) {
-        new_key <- paste(key_prefix, nm[i], sep = sep)
+        if (nzchar(key_prefix)) {
+          new_key <- paste(key_prefix, nm[i], sep = sep)
+        } else {
+          new_key <- nm[i]
+        }
         recurse(x[[i]], new_key)
       }
     } else {

R/btw_client_app.R

@@ -118,6 +118,7 @@ btw_app_from_client <- function(client, messages = list(), ...) {
         class = "btn-close",
         style = "position: fixed; top: 6px; right: 6px;"
       ),
+      class = "bslib-page-dashboard",
       btw_title(FALSE),
       shinychat::chat_mod_ui(
         "chat",
@@ -137,6 +138,13 @@ btw_app_from_client <- function(client, messages = list(), ...) {
         .sidebar-collapsed > .main > main .sidebar-title { display: block; }
         .bslib-sidebar-layout.sidebar-collapsed>.collapse-toggle { top: 1.8rem; }
         .bslib-page-main { gap: 0.5rem; }
+        aside#tools_sidebar {
+            box-shadow: 2px 2px 5px rgba(var(--bs-emphasis-color-rgb), 10%);
+        }
+        shiny-chat-message .message-icon {
+          background-color: var(--bs-white);
+          box-shadow: 2px 2px 5px rgba(var(--bs-emphasis-color-rgb), 10%);
+        }
       "
         )),
       )
@@ -215,13 +223,16 @@ btw_app_from_client <- function(client, messages = list(), ...) {
       if (!length(selected_tools())) {
         client$set_tools(list())
       } else {
-        .btw_tools <- keep(btw_tools(), function(tool) {
-          tool@name %in% selected_tools()
-        })
-        .other_tools <- keep(other_tools, function(tool) {
+        sel_btw_tools <- btw_tools(
+          intersect(names(.btw_tools), selected_tools())
+        )
+        sel_other_tools <- keep(other_tools, function(tool) {
           tool@name %in% selected_tools()
         })
-        client$set_tools(c(.btw_tools, other_tools))
+        sel_tools <- c(sel_btw_tools, sel_other_tools)
+        # tool_names <- map_chr(tools, S7::prop, "name")
+        # cli::cli_inform("Setting {.field client} tools to: {.val {tool_names}}")
+        client$set_tools(sel_tools)
       }
     })
 
@@ -269,6 +280,10 @@ btw_app_from_client <- function(client, messages = list(), ...) {
     save.interface = old_save
   ))
 
+  if (identical(Sys.getenv("BTW_IN_TESTING"), "true")) {
+    return(list(ui = ui, server = server))
+  }
+
   app <- shiny::shinyApp(ui, server, ...)
   if (getOption("btw.app.in_addin", FALSE)) {
     shiny::runApp(app, launch.browser = function(url) {
@@ -597,10 +612,12 @@ app_tool_group_choice_input <- function(
     group,
     "docs" = shiny::span(label_icon, "Documentation"),
     "env" = shiny::span(label_icon, "Environment"),
+    "eval" = shiny::span(label_icon, "Code Evaluation"),
     "files" = shiny::span(label_icon, "Files"),
     "git" = shiny::span(label_icon, "Git"),
     "github" = shiny::span(label_icon, "GitHub"),
     "ide" = shiny::span(label_icon, "IDE"),
+    "run" = shiny::span(label_icon, "Run Code"),
     "search" = shiny::span(label_icon, "Search"),
     "session" = shiny::span(label_icon, "Session Info"),
     "web" = shiny::span(label_icon, "Web Tools"),