ToolCallMessage¶

Tool invocation request from the LLM.

Class: `RobotLab::ToolCallMessage`¶

message = ToolCallMessage.new(
  id: "call_abc123",
  name: "get_weather",
  input: { city: "New York", units: "fahrenheit" }
)

Constructor¶

ToolCallMessage.new(id:, name:, input:)

Parameters:

Name	Type	Description
`id`	`String`	Unique call identifier
`name`	`String`	Tool name
`input`	`Hash`	Tool parameters

Attributes¶

id¶

message.id  # => String

Unique identifier for this tool call. Used to match with ToolResultMessage.

name¶

message.name  # => String

Name of the tool being invoked.

input¶

message.input  # => Hash

Parameters passed to the tool.

role¶

message.role  # => :assistant

Always returns :assistant (the LLM initiates tool calls).

Methods¶

to_h¶

message.to_h  # => Hash

Hash representation.

Returns:

{
  role: :assistant,
  tool_call: {
    id: "call_abc123",
    name: "get_weather",
    input: { city: "New York", units: "fahrenheit" }
  }
}

to_json¶

message.to_json  # => String

JSON representation.

Examples¶

Basic Tool Call¶

call = ToolCallMessage.new(
  id: "call_1",
  name: "search_orders",
  input: { user_id: "123", status: "pending" }
)

Processing Tool Calls¶

result.output.each do |msg|
  case msg
  when ToolCallMessage
    puts "Tool called: #{msg.name}"
    puts "Parameters: #{msg.input.inspect}"
  when ToolResultMessage
    puts "Result for #{msg.id}: #{msg.result}"
  end
end

In Tool Execution Flow¶

# LLM returns a tool call
tool_call = ToolCallMessage.new(
  id: "call_weather_1",
  name: "get_weather",
  input: { city: "Seattle" }
)

# Tool is executed
result = tool.call(tool_call.input, state: state)

# Result is recorded
tool_result = ToolResultMessage.new(
  id: tool_call.id,
  result: result
)