Tabulate a tool call's arguments for review

Builds a data frame with one row per argument, pairing the value the model proposed with the tool's declared type and description. Nested objects are flattened to dotted names such as plan.treatment. Use it in a can_use_tool callback, a PreToolUse hook or an approval screen to show a person a table instead of dput() output.

Permission callbacks and PreToolUse hooks get the tool's declared arguments as context$tool_arguments. The table is for display only; validate inputs in the tool itself.

Usage

tool_input_review(tool_input, tool_arguments = NULL)

Arguments

tool_input

Named list of proposed arguments, as passed to permission callbacks and hooks, or the request$tool_input of a pending approval from approval_read().

tool_arguments

The tool's declared arguments: an ellmer TypeObject such as context$tool_arguments or tool@arguments, or NULL if you don't have one.

Value

A data frame with character columns argument, type, description and value, and logical columns required and declared. Input fields the tool doesn't declare are kept with declared = FALSE; declared fields missing from the input have value NA. The "paths" attribute holds each row's path as a character vector, which stays exact when a name itself contains a dot.

Examples

arguments <- ellmer::type_object(
  city = ellmer::type_string("City to forecast"),
  days = ellmer::type_integer("Number of days", required = FALSE)
)
tool_input_review(list(city = "Oslo", days = 3L), arguments)
#>   argument    type required declared      description value
#> 1     city  string     TRUE     TRUE City to forecast  Oslo
#> 2     days integer    FALSE     TRUE   Number of days     3