mirror of
https://github.com/OpenVoiceOS/OpenVoiceOS
synced 2024-12-16 10:58:57 +01:00
400 lines
15 KiB
Python
Executable File
400 lines
15 KiB
Python
Executable File
# Copyright 2017 Mycroft AI Inc.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
#
|
|
from alsaaudio import Mixer, mixers as alsa_mixers
|
|
from os.path import dirname, join
|
|
|
|
from adapt.intent import IntentBuilder
|
|
from mycroft.audio import wait_while_speaking
|
|
from mycroft.messagebus.message import Message
|
|
from mycroft.skills.core import MycroftSkill, intent_handler
|
|
from mycroft.util import play_wav
|
|
from mycroft.util.parse import extract_number
|
|
|
|
|
|
ALSA_PLATFORMS = ['MycroftOS', 'mycroft_mark_1', 'picroft', 'unknown']
|
|
|
|
|
|
class VolumeSkill(MycroftSkill):
|
|
"""
|
|
Control the audio volume for the Mycroft system
|
|
|
|
Terminology:
|
|
"Level" = Mycroft volume levels, from 0 to 10
|
|
"Volume" = ALSA mixer setting, from 0 to 100
|
|
"""
|
|
|
|
MIN_LEVEL = 0
|
|
MAX_LEVEL = 10
|
|
|
|
# TODO: Translation layer (have to match word in Level.voc)
|
|
VOLUME_WORDS = {
|
|
'loud': 9,
|
|
'normal': 6,
|
|
'quiet': 3
|
|
}
|
|
|
|
def __init__(self):
|
|
super(VolumeSkill, self).__init__("VolumeSkill")
|
|
self.settings["default_level"] = 6 # can be 0 (off) to 10 (max)
|
|
self.settings["min_volume"] = 0 # can be 0 to 100
|
|
if self.config_core['enclosure'].get('platform') == 'mycroft_mark_1':
|
|
self.settings["max_volume"] = 83 # can be 0 to 83
|
|
else:
|
|
self.settings["max_volume"] = 100 # can be 0 to 100
|
|
self.volume_sound = join(dirname(__file__), "blop-mark-diangelo.wav")
|
|
self.vol_before_mute = None
|
|
self._mixer = None
|
|
|
|
def _clear_mixer(self):
|
|
"""For Unknown platforms reinstantiate the mixer.
|
|
|
|
For mycroft_mark_1 do not reinstantiate the mixer.
|
|
"""
|
|
platform = self.config_core['enclosure'].get('platform', 'unknown')
|
|
if platform != 'mycroft_mark_1':
|
|
self._mixer = None
|
|
|
|
def _get_mixer(self):
|
|
self.log.debug('Finding Alsa Mixer for control...')
|
|
mixer = None
|
|
try:
|
|
# If there are only 1 mixer use that one
|
|
mixers = alsa_mixers()
|
|
if len(mixers) == 1:
|
|
mixer = Mixer(mixers[0])
|
|
elif 'Master' in mixers:
|
|
# Try using the default mixer (Master)
|
|
mixer = Mixer('Master')
|
|
elif 'PCM' in mixers:
|
|
# PCM is another common one
|
|
mixer = Mixer('PCM')
|
|
elif 'Digital' in mixers:
|
|
# My mixer is called 'Digital' (JustBoom DAC)
|
|
mixer = Mixer('Digital')
|
|
else:
|
|
# should be equivalent to 'Master'
|
|
mixer = Mixer()
|
|
except Exception:
|
|
# Retry instanciating the mixer with the built-in default
|
|
try:
|
|
mixer = Mixer()
|
|
except Exception as e:
|
|
self.log.error('Couldn\'t allocate mixer, {}'.format(repr(e)))
|
|
self._mixer = mixer
|
|
return mixer
|
|
|
|
def initialize(self):
|
|
# Register handlers to detect percentages as reported by STT
|
|
for i in range(101): # numbers 0 to 100
|
|
self.register_vocabulary(str(i) + '%', 'Percent')
|
|
|
|
# Register handlers for messagebus events
|
|
self.add_event('mycroft.volume.increase',
|
|
self.handle_increase_volume)
|
|
self.add_event('mycroft.volume.decrease',
|
|
self.handle_decrease_volume)
|
|
self.add_event('mycroft.volume.mute',
|
|
self.handle_mute_volume)
|
|
self.add_event('mycroft.volume.unmute',
|
|
self.handle_unmute_volume)
|
|
self.add_event('recognizer_loop:record_begin',
|
|
self.duck)
|
|
self.add_event('recognizer_loop:record_end',
|
|
self.unduck)
|
|
|
|
self.vol_before_mute = self.__get_system_volume()
|
|
|
|
@property
|
|
def mixer(self):
|
|
platform = self.config_core['enclosure'].get('platform', 'unknown')
|
|
if platform in ALSA_PLATFORMS:
|
|
return self._mixer or self._get_mixer()
|
|
else:
|
|
return None
|
|
|
|
def _setvolume(self, vol, emit=True):
|
|
# Update ALSA
|
|
if self.mixer:
|
|
self.log.debug(vol)
|
|
self.mixer.setvolume(vol)
|
|
# TODO: Remove this and control volume at the Enclosure level in
|
|
# response to the mycroft.volume.set message.
|
|
|
|
if emit:
|
|
# Notify non-ALSA systems of volume change
|
|
self.bus.emit(Message('mycroft.volume.set',
|
|
data={"percent": vol/100.0}))
|
|
|
|
# Change Volume to X (Number 0 to) Intent Handlers
|
|
@intent_handler(IntentBuilder("SetVolume").require("Volume")
|
|
.optionally("Increase").optionally("Decrease")
|
|
.optionally("To").require("Level"))
|
|
def handle_set_volume(self, message):
|
|
self._clear_mixer()
|
|
default_vol = self.__get_system_volume(50)
|
|
|
|
level = self.__get_volume_level(message, default_vol)
|
|
self._setvolume(self.__level_to_volume(level))
|
|
if level == self.MAX_LEVEL:
|
|
self.speak_dialog('max.volume')
|
|
else:
|
|
self.speak_dialog('set.volume', data={'volume': level})
|
|
|
|
# Set Volume Percent Intent Handlers
|
|
@intent_handler(IntentBuilder("SetVolumePercent").require("Volume")
|
|
.optionally("Increase").optionally("Decrease")
|
|
.optionally("To").require("Percent"))
|
|
def handle_set_volume_percent(self, message):
|
|
self._clear_mixer()
|
|
percent = extract_number(message.data['utterance'].replace('%', ''))
|
|
percent = int(percent)
|
|
self._setvolume(percent)
|
|
self.speak_dialog('set.volume.percent', data={'level': percent})
|
|
|
|
# Volume Status Intent Handlers
|
|
@intent_handler(IntentBuilder("QueryVolume").optionally("Query")
|
|
.require("Volume"))
|
|
def handle_query_volume(self, message):
|
|
self._clear_mixer()
|
|
level = self.__volume_to_level(self.__get_system_volume(0, show=True))
|
|
self.speak_dialog('volume.is', data={'volume': round(level)})
|
|
|
|
@intent_handler(IntentBuilder("QueryVolumePhrase").require("QueryPhrase")
|
|
.optionally("Volume"))
|
|
def handle_query_volume_phrase(self, message):
|
|
self.handle_query_volume(message)
|
|
|
|
def __communicate_volume_change(self, message, dialog, code, changed):
|
|
play_sound = message.data.get('play_sound', False)
|
|
if play_sound:
|
|
if changed:
|
|
play_wav(self.volume_sound)
|
|
else:
|
|
if (not changed) and (code != 0):
|
|
self.speak_dialog('already.max.volume', data={'volume': code})
|
|
|
|
# Increase Volume Intent Handlers
|
|
@intent_handler(IntentBuilder("IncreaseVolume").require("Volume")
|
|
.require("Increase"))
|
|
def handle_increase_volume(self, message):
|
|
self.__communicate_volume_change(message, 'increase.volume',
|
|
*self.__update_volume(+1))
|
|
|
|
@intent_handler(IntentBuilder("IncreaseVolumeSet").require("Set")
|
|
.optionally("Volume").require("Increase"))
|
|
def handle_increase_volume_set(self, message):
|
|
self._clear_mixer()
|
|
self.handle_increase_volume(message)
|
|
|
|
@intent_handler(IntentBuilder("IncreaseVolumePhrase")
|
|
.require("IncreasePhrase"))
|
|
def handle_increase_volume_phrase(self, message):
|
|
self._clear_mixer()
|
|
self.handle_increase_volume(message)
|
|
|
|
# Decrease Volume Intent Handlers
|
|
@intent_handler(IntentBuilder("DecreaseVolume").require("Volume")
|
|
.require("Decrease"))
|
|
def handle_decrease_volume(self, message):
|
|
self.__communicate_volume_change(message, 'decrease.volume',
|
|
*self.__update_volume(-1))
|
|
|
|
@intent_handler(IntentBuilder("DecreaseVolumeSet").require("Set")
|
|
.optionally("Volume").require("Decrease"))
|
|
def handle_decrease_volume_set(self, message):
|
|
self.handle_decrease_volume(message)
|
|
|
|
@intent_handler(IntentBuilder("DecreaseVolumePhrase")
|
|
.require("DecreasePhrase"))
|
|
def handle_decrease_volume_phrase(self, message):
|
|
self.handle_decrease_volume(message)
|
|
|
|
# Maximum Volume Intent Handlers
|
|
@intent_handler(IntentBuilder("MaxVolume").optionally("Set")
|
|
.require("Volume").optionally("Increase")
|
|
.require("MaxVolume"))
|
|
def handle_max_volume(self, message):
|
|
self._clear_mixer()
|
|
self._setvolume(self.settings["max_volume"])
|
|
speak_message = message.data.get('speak_message', True)
|
|
if speak_message:
|
|
self.speak_dialog('max.volume')
|
|
wait_while_speaking()
|
|
self.bus.emit(Message('mycroft.volume.duck'))
|
|
|
|
@intent_handler(IntentBuilder("MaxVolumeIncreaseMax")
|
|
.require("MaxVolumePhrase").optionally("Volume")
|
|
.require("Increase").optionally("MaxVolume"))
|
|
def handle_max_volume_increase_to_max(self, message):
|
|
self.handle_max_volume(message)
|
|
|
|
def duck(self, message):
|
|
self._clear_mixer()
|
|
if self.settings.get('ducking', True):
|
|
self._mute_volume()
|
|
|
|
def unduck(self, message):
|
|
self._clear_mixer()
|
|
if self.settings.get('ducking', True):
|
|
self._unmute_volume()
|
|
|
|
def _mute_volume(self, message=None, speak=False):
|
|
self.log.debug('MUTING!')
|
|
self.vol_before_mute = self.__get_system_volume()
|
|
self.log.debug(self.vol_before_mute)
|
|
if speak:
|
|
self.speak_dialog('mute.volume')
|
|
wait_while_speaking()
|
|
self._setvolume(0, emit=False)
|
|
self.bus.emit(Message('mycroft.volume.duck'))
|
|
|
|
# Mute Volume Intent Handlers
|
|
@intent_handler(IntentBuilder("MuteVolume").require(
|
|
"Volume").require("Mute"))
|
|
def handle_mute_volume(self, message):
|
|
self._clear_mixer()
|
|
self._mute_volume(speak=message.data.get('speak_message', True))
|
|
|
|
def _unmute_volume(self, message=None, speak=False):
|
|
if self.vol_before_mute is None:
|
|
vol = self.__level_to_volume(self.settings["default_level"])
|
|
else:
|
|
vol = self.vol_before_mute
|
|
self.vol_before_mute = None
|
|
|
|
self._setvolume(vol, emit=False)
|
|
self.bus.emit(Message('mycroft.volume.unduck'))
|
|
|
|
if speak:
|
|
self.speak_dialog('reset.volume',
|
|
data={'volume':
|
|
self.settings["default_level"]})
|
|
|
|
# Unmute/Reset Volume Intent Handlers
|
|
@intent_handler(IntentBuilder("UnmuteVolume").require("Volume")
|
|
.require("Unmute"))
|
|
def handle_unmute_volume(self, message):
|
|
self._clear_mixer()
|
|
self._unmute_volume(speak=message.data.get('speak_message', True))
|
|
|
|
def __volume_to_level(self, volume):
|
|
"""
|
|
Convert a 'volume' to a 'level'
|
|
|
|
Args:
|
|
volume (int): min_volume..max_volume
|
|
Returns:
|
|
int: the equivalent level
|
|
"""
|
|
range = self.MAX_LEVEL - self.MIN_LEVEL
|
|
min_vol = self.settings["min_volume"]
|
|
max_vol = self.settings["max_volume"]
|
|
prop = float(volume - min_vol) / max_vol
|
|
level = int(round(self.MIN_LEVEL + range * prop))
|
|
if level > self.MAX_LEVEL:
|
|
level = self.MAX_LEVEL
|
|
elif level < self.MIN_LEVEL:
|
|
level = self.MIN_LEVEL
|
|
return level
|
|
|
|
def __level_to_volume(self, level):
|
|
"""
|
|
Convert a 'level' to a 'volume'
|
|
|
|
Args:
|
|
level (int): 0..MAX_LEVEL
|
|
Returns:
|
|
int: the equivalent volume
|
|
"""
|
|
range = self.settings["max_volume"] - self.settings["min_volume"]
|
|
prop = float(level) / self.MAX_LEVEL
|
|
volume = int(round(self.settings["min_volume"] + int(range) * prop))
|
|
|
|
return volume
|
|
|
|
@staticmethod
|
|
def __bound_level(level):
|
|
if level > VolumeSkill.MAX_LEVEL:
|
|
level = VolumeSkill.MAX_LEVEL
|
|
elif level < VolumeSkill.MIN_LEVEL:
|
|
level = VolumeSkill.MIN_LEVEL
|
|
return level
|
|
|
|
def __update_volume(self, change=0):
|
|
"""
|
|
Attempt to change audio level
|
|
|
|
Args:
|
|
change (int): +1 or -1; the step to change by
|
|
|
|
Returns: tuple(new level code int(0..10),
|
|
whether level changed (bool))
|
|
"""
|
|
old_level = self.__volume_to_level(self.__get_system_volume(0))
|
|
new_level = self.__bound_level(old_level + change)
|
|
self.enclosure.eyes_volume(new_level)
|
|
self._setvolume(self.__level_to_volume(new_level))
|
|
return new_level, new_level != old_level
|
|
|
|
def __get_system_volume(self, default=50, show=False):
|
|
""" Get volume, either from mixer or ask on messagebus.
|
|
|
|
The show parameter should only be True when a user is requesting
|
|
the volume and not the system.
|
|
TODO: Remove usage of Mixer and move that stuff to enclosure.
|
|
"""
|
|
vol = default
|
|
if self.mixer:
|
|
vol = min(self.mixer.getvolume()[0], 100)
|
|
self.log.debug('Volume before mute: {}'.format(vol))
|
|
else:
|
|
vol_msg = self.bus.wait_for_response(
|
|
Message("mycroft.volume.get", {'show': show}))
|
|
if vol_msg:
|
|
vol = int(vol_msg.data["percent"] * 100)
|
|
|
|
return vol
|
|
|
|
def __get_volume_level(self, message, default=None):
|
|
""" Retrievs volume from message. """
|
|
level_str = message.data.get('Level', default)
|
|
level = self.settings["default_level"]
|
|
|
|
try:
|
|
level = self.VOLUME_WORDS[level_str]
|
|
except KeyError:
|
|
try:
|
|
level = int(extract_number(level_str))
|
|
if (level == self.MAX_LEVEL + 1):
|
|
# Assume that user meant max volume
|
|
level = self.MAX_LEVEL
|
|
elif (level > self.MAX_LEVEL):
|
|
# Guess that the user said something like 100 percent
|
|
# so convert that into a level value
|
|
level = self.MAX_LEVEL * level/100
|
|
except ValueError:
|
|
pass
|
|
|
|
level = self.__bound_level(level)
|
|
return level
|
|
|
|
def shutdown(self):
|
|
if self.vol_before_mute is not None:
|
|
self._unmute_volume()
|
|
|
|
|
|
def create_skill():
|
|
return VolumeSkill()
|