Collecting structured input: get_response with a validator

In a nutshell

You build an OVOSSkill intent handler that collects one specific, validated value, using get_response(validator=..., on_fail=..., num_retries=...) with a bounded retry count.

When you'd want this: an intent needs one specific piece of information, such as a number in a known range, not a free-form sentence. The user should be re-prompted on a bad answer, a limited number of times, and the skill should give up gracefully rather than loop forever.

from ovos_workshop.decorators import intent_handler
from ovos_workshop.skills import OVOSSkill


class ThermostatSkill(OVOSSkill):
    @intent_handler("set_temperature.intent")
    def handle_set_temperature(self, message):
        def in_range(utterance: str) -> bool:
            try:
                return 10 <= int(utterance) <= 30
            except ValueError:
                return False

        response = self.get_response("ask_temperature",
                                     validator=in_range,
                                     on_fail="temperature_out_of_range",
                                     num_retries=2)
        if response is None:
            self.speak_dialog("temperature_not_set")
            return

        self.settings["target_temperature"] = int(response)
        self.speak_dialog("temperature_set", {"temp": response})

Moving parts

• get_response(dialog='', data=None, validator=None, on_fail=None, num_retries=-1, message=None, wait=True) speaks dialog, listens for a reply, and, if validator is given, keeps re-asking until the reply passes it or the retries run out. validator takes the raw utterance string and returns True/False; in_range above rejects anything that isn't an integer between 10 and 30.
• on_fail is what gets spoken on a failed validation, before listening again. It can be a dialog name/string, or a (str) -> str callable that builds a message from the failing utterance.
• num_retries bounds the re-prompt loop. 2 above means the user gets up to three attempts total (the original ask plus two retries) before get_response gives up. -1, the default, retries indefinitely, but a silent (no-answer) timeout with -1 still only retries once. Cap num_retries deliberately whenever a bad or silent answer should end the interaction instead of looping.
• get_response returns the matched utterance as a str on success, or None if no valid response was ever captured, whether because retries ran out, the user said "cancel", or nobody answered. handle_set_temperature above checks for None and speaks a distinct "gave up" dialog rather than assuming response is always usable. See Get User Response for get_response alongside ask_yesno and ask_selection.

Full production example

ovos-skill-mark1-ctrl's handle_custom_eye_color chains three validated get_response calls back to back, one each for red, green, and blue (0-255), each with its own on_fail dialog and num_retries=2, then falls through to ask_yesno to offer saving the result as the default eye color.

---

Read next: Skill Cookbook Related: Get User Response · Continuous conversation · ovos-skill-mark1-ctrl