Class: Tools::Base Abstract

Inherits:

Object

• Object
• Tools::Base

Defined in:

lib/tools/base.rb

Overview

This class is abstract.

Subclass and implement Base.tool_name, Base.description, Base.input_schema, and #execute

Abstract base class for all Anima tools. Subclasses must implement the class-level schema methods and the instance-level #execute method.

Examples:

Implementing a tool

class Tools::Echo < Tools::Base
  def self.tool_name = "echo"
  def self.description = "Echoes input back"
  def self.input_schema
    {type: "object", properties: {text: {type: "string"}}, required: ["text"]}
  end

  def execute(input)
    input["text"]
  end
end

Direct Known Subclasses

AnalyticalBrain::Tools::ActivateSkill, AnalyticalBrain::Tools::AssignNickname, AnalyticalBrain::Tools::DeactivateSkill, AnalyticalBrain::Tools::DeactivateWorkflow, AnalyticalBrain::Tools::EverythingIsReady, AnalyticalBrain::Tools::FinishGoal, AnalyticalBrain::Tools::ReadWorkflow, AnalyticalBrain::Tools::RenameSession, AnalyticalBrain::Tools::SetGoal, AnalyticalBrain::Tools::UpdateGoal, Mneme::Tools::AttachMessagesToGoals, Mneme::Tools::EverythingOk, Mneme::Tools::SaveSnapshot, Bash, Edit, OpenIssue, Read, Remember, SpawnSpecialist, SpawnSubagent, Think, WebGet, Write

Class Method Summary

• .description ⇒ String

  Human-readable description for the LLM.

• .input_schema ⇒ Hash

  JSON Schema describing the tool’s input parameters.

• .schema ⇒ Hash

  Builds the schema hash expected by the Anthropic tools API.

• .tool_name ⇒ String

  Unique tool identifier sent to the LLM.

Instance Method Summary

• #execute(input) ⇒ String, Hash

  Execute the tool with the given input.

• #initialize ⇒ Base constructor

  Accepts and discards context keywords so that the Registry can pass shared dependencies (e.g. shell_session) to any tool uniformly.

Constructor Details

#initialize ⇒ Base

Accepts and discards context keywords so that the Registry can pass shared dependencies (e.g. shell_session) to any tool uniformly. Subclasses that need specific context should override with named kwargs.

# File 'lib/tools/base.rb', line 49

def initialize(**) = nil

Class Method Details

.description ⇒ String

Returns human-readable description for the LLM.

Returns:

• (String) —

  human-readable description for the LLM

Raises:

• (NotImplementedError)

# File 'lib/tools/base.rb', line 30

def description
  raise NotImplementedError, "#{self} must implement .description"
end

.input_schema ⇒ Hash

Returns JSON Schema describing the tool’s input parameters.

Returns:

• (Hash) —

  JSON Schema describing the tool’s input parameters

Raises:

• (NotImplementedError)

# File 'lib/tools/base.rb', line 35

def input_schema
  raise NotImplementedError, "#{self} must implement .input_schema"
end

.schema ⇒ Hash

Builds the schema hash expected by the Anthropic tools API.

Returns:

• (Hash) —

  with :name, :description, and :input_schema keys

# File 'lib/tools/base.rb', line 41

def schema
  {name: tool_name, description: description, input_schema: input_schema}
end

.tool_name ⇒ String

Returns unique tool identifier sent to the LLM.

Returns:

• (String) —

  unique tool identifier sent to the LLM

Raises:

• (NotImplementedError)

# File 'lib/tools/base.rb', line 25

def tool_name
  raise NotImplementedError, "#{self} must implement .tool_name"
end

Instance Method Details

#execute(input) ⇒ String, Hash

Execute the tool with the given input.

Parameters:

• input (Hash) —

  parsed input matching input_schema. May include a “timeout” key (seconds) for tools that support per-call timeout overrides.

Returns:

• (String, Hash) —

  result content; Hash with :error key signals failure

Raises:

• (NotImplementedError)

# File 'lib/tools/base.rb', line 55

def execute(input)
  raise NotImplementedError, "#{self.class} must implement #execute"
end