Start a WEAR loop data exploration session

Launches an interactive data exploration session using the Write/Execute/Analyze/Regroup (WEAR) loop pattern. This is a dedicated exploration mode – it is separate from a normal codeagent_app() or codeagent_console() session. The difference:

Usage

wear_explore(data = NULL, client = NULL, mode = c("repl", "shiny"), ...)

Arguments

data

Named list, environment, or NULL. Data.frames to make available for exploration. If NULL, uses objects in .GlobalEnv. A named list is converted to an environment automatically.

client

A CodeagentClient (from codeagent_client()). If NULL, one is built from ~/.codeagent/settings.json with permission_mode = "bypass".

mode

Character. "repl" (default) starts an interactive CLI session; "shiny" launches the Shiny app.

...

Passed to codeagent_console() or codeagent_app().

Value

Invisibly the CodeagentClient (useful for post-session inspection or calling generate_wear_report() manually).

Details

Normal session	WEAR session (wear_explore())
No data registered by default	data= argument registers named data.frames
No GenerateReport tool	/report exports session to .qmd
No WEAR system prompt	Agent instructed to end each turn with Next steps
General-purpose tools	ExploreData tool added (read-only, sandboxed)

The ExploreData and GenerateReport tools are not registered in the standard agent loop (codeagent_app()) – use wear_explore() to enter exploration mode explicitly.

See also

generate_wear_report() to export the session to a Quarto document.

Examples

if (FALSE) { # \dontrun{
# Explore mtcars from the CLI (requires a configured LLM client)
wear_explore(data = list(mtcars = mtcars))

# Shiny UI with multiple data.frames
wear_explore(
  data = list(sales = sales_df, products = products_df),
  mode = "shiny"
)

# Capture the client to export the report afterwards
client <- wear_explore(data = list(mtcars = mtcars))
generate_wear_report(client, title = "Motor Trend Analysis")
} # }