Create validation tool for job execution — mcpflow_job

Creates a callback function that validates tool calls against workflow job definitions. The validation tool tracks which jobs have been completed and warns (or rejects) when tools are called out of expected order or when job dependencies are not met.

Usage

mcpflow_job_validator(workflow, strict = FALSE, verbose = TRUE)

Arguments

workflow: A MCP workflow object
strict: Logical. If TRUE, reject out-of-order tool calls via tool_reject. If FALSE (default), only warn via cli_warn.
verbose: Logical. If TRUE (default), print progress messages when jobs and steps are detected.

Value

A list with two callback functions:

on_tool_request: Callback for chat$on_tool_request()
on_tool_result: Callback for chat$on_tool_result()
state: An environment containing execution state (for inspection)

Details

The validation tool maintains state about which jobs have been started/completed based on tool calls observed. It checks:

Whether the tool being called matches expected job steps
Whether job dependencies (needs) are satisfied
Step execution order within jobs

Examples


wf <- mcpflow_read("ravepipeline::rave_pipeline_class_guide")
validator <- mcpflow_job_validator(wf, strict = FALSE)

if (FALSE) { # \dontrun{
chat <- ellmer::chat_ollama()
chat$on_tool_request(validator$on_tool_request)
chat$on_tool_result(validator$on_tool_result)
} # }