StateMachineInstruction

StateMachineInstruction defines the single-state instruction used when each input maps to one output state. It maps one or more input TokenSets to a state string selected from your samples.

Key characteristics:

• Output is created automatically with a single State token
• Final token is always <NON>
• Samples provide a state string instead of a response snippet

StateMachineInstruction Parameters

class StateMachineInstruction:
    def __init__(self, input: StateMachineInput, states: List[str]):

• input: A StateMachineInput instance defining the input TokenSets
• states: A list of allowed state names for the instruction output

Creating a StateMachineInstruction

question_tokenset = mtp.TokenSet(tokens=(question, context))

state_machine_input = mtp.StateMachineInput(
    tokensets=[question_tokenset]
)

state_machine_instruction = mtp.StateMachineInstruction(
    input=state_machine_input,
    states=["WHERE"]
)

state_machine_instruction.add_sample(
    input_snippets=["Where are you from?"],
    state="WHERE"
)

The list of possible states is derived from these samples and stored as the single-state output metadata when state_machine=True. For the multi-state form, see MultiClassifierInstruction. For shared input structure, see StateMachineInput.