Nous.Tools.ReActTools (nous v0.16.1)

View Source

Built-in tools for ReAct (Reasoning and Acting) agents.

These tools provide structured planning, note-taking, todo management, and explicit task completion for ReAct agents.

Tools Included

Usage

agent = ReActAgent.new("lmstudio:qwen3-vl-4b-thinking-mlx",
  instructions: "Solve complex problems step by step"
)

The agent will automatically have access to all ReAct tools.

Summary

Functions

Add a task to the todo list.

Mark a todo item as complete.

Provide the final answer to complete the task.

List all current todos with their status.

Record an observation or insight during the reasoning process.

Create a structured plan for solving a task.

Functions

add_todo(ctx, args)

Add a task to the todo list.

Use this to break down complex problems into manageable subtasks. Track what needs to be done and maintain focus.

Arguments

  • item - Description of the todo item (string)
  • priority - Optional priority: "high", "medium", "low" (default: "medium")

Returns

Confirmation with current todo count.

complete_todo(ctx, args)

Mark a todo item as complete.

Arguments

  • id - The todo ID to complete (integer)

OR

  • item - Description matching the todo item (string)

Returns

Confirmation message.

final_answer(ctx, map)

Provide the final answer to complete the task.

This tool MUST be called to signal task completion. The agent should only use this after gathering all necessary information and solving the problem.

Arguments

  • answer - The final answer or solution (string)

Returns

The final answer wrapped with completion metadata.

list_todos(ctx, args \\ %{})

List all current todos with their status.

Shows pending and completed tasks to help track progress.

Returns

Formatted list of todos.

note(ctx, map)

Record an observation or insight during the reasoning process.

Use this to document important findings, intermediate results, or thoughts that will inform future actions.

Arguments

  • content - The note content (string)

Returns

Confirmation message.

plan(ctx, args)

Create a structured plan for solving a task.

Helps the agent organize its approach by identifying:

  • Known facts
  • Facts to look up
  • Facts to derive/calculate
  • Step-by-step action plan

Arguments

  • task - The task or problem to plan for (string)
  • Can also be called with no arguments to plan the current conversation

Returns

A confirmation message. The plan is stored in context for reference.