Jido.BehaviorTree.Skill (Jido Behavior Tree v1.0.0)

View Source

A behavior tree skill that can be converted to AI-compatible tool definitions.

This module allows behavior trees to be exposed as skills that can be executed by AI systems. The behavior tree is wrapped in a skill interface that provides parameter validation, execution, and result formatting.

Features

  • Convert behavior trees to LLM-compatible tool definitions
  • Parameter validation and type checking
  • Automatic blackboard management
  • Execution result formatting
  • Error handling and reporting

Example Usage

# Create a behavior tree
tree = Jido.BehaviorTree.Tree.new(root_node)

# Create a skill from the tree
skill = Jido.BehaviorTree.Skill.new(
  "process_user_data",
  tree,
  "Processes user data through a behavior tree",
  schema: [
    user_id: [type: :integer, required: true],
    data: [type: :map, required: true]
  ]
)

# Convert to tool format
tool_def = skill.to_tool()

# Execute the skill
{:ok, result} = skill.run(%{user_id: 123, data: %{name: "John"}}, %{})

Summary

Types

t()

Functions

new(name, tree, description \\ "", opts \\ [])

Creates a new behavior tree skill.

run(skill, params, context)

Executes the behavior tree skill with the given parameters.

schema()

Returns the Zoi schema for this module

to_tool(skill)

Converts the skill to an AI-compatible tool definition.

Types

t()

@type t() :: %Jido.BehaviorTree.Skill{
  auto_mode: boolean(),
  description: binary(),
  interval: integer(),
  name: binary(),
  output_schema: any(),
  schema: any(),
  timeout: integer(),
  tree: any()
}

Functions

new(name, tree, description \\ "", opts \\ [])

@spec new(String.t(), Jido.BehaviorTree.Tree.t(), String.t(), keyword()) :: t()

Creates a new behavior tree skill.

Options

  • :schema - NimbleOptions schema for input validation (default: [])
  • :output_schema - NimbleOptions schema for output validation (default: [])
  • :timeout - Execution timeout in milliseconds (default: 30_000)
  • :auto_mode - Whether to run in automatic mode (default: false)
  • :interval - Tick interval for auto mode in milliseconds (default: 1000)

Examples

skill = Jido.BehaviorTree.Skill.new(
  "data_processor",
  tree,
  "Processes data using behavior tree logic",
  schema: [
    input_data: [type: :map, required: true]
  ],
  timeout: 10_000
)

run(skill, params, context)

@spec run(t(), map(), map()) :: {:ok, map()} | {:error, term()}

Executes the behavior tree skill with the given parameters.

This function starts a behavior tree agent, executes the tree with the provided parameters in the blackboard, and returns the final result.

Parameters

  • params - Input parameters (will be validated against schema)
  • context - Execution context (currently unused but reserved for future use)

Returns

  • {:ok, result} - Successful execution with result map
  • {:error, reason} - Execution failed with error reason

Examples

{:ok, result} = Jido.BehaviorTree.Skill.run(skill, %{input_data: %{id: 1}}, %{})

schema()

Returns the Zoi schema for this module

to_tool(skill)

@spec to_tool(t()) :: map()

Converts the skill to an AI-compatible tool definition.

Returns a map that can be used with LLM function calling systems like OpenAI's function calling.

Examples

tool_def = skill.to_tool()
# %{
#   "name" => "data_processor",
#   "description" => "Processes data using behavior tree logic",
#   "parameters" => %{
#     "type" => "object",
#     "properties" => %{...},
#     "required" => [...]
#   }
# }