This commit officially switches the mycroft-core repository from
GPLv3.0 licensing to Apache 2.0. All dependencies on GPL'ed code
have been removed and we have contacted all previous contributors
with still-existing code in the repository to agree to this change.
Going forward, all contributors will sign a Contributor License
Agreement (CLA) by visiting https://mycroft.ai/cla, then they will
be included in the Mycroft Project's overall Contributor list,
found at: https://github.com/MycroftAI/contributors. This cleanly
protects the project, the contributor and all who use the technology
to build upon.
Futher discussion can be found at this blog post:
https://mycroft.ai/blog/right-license/
This commit also removes all __author__="" from the code. These
lines are painful to maintain and the etiquette surrounding their
maintainence is unclear. Do you remove a name from the list if the
last line of code the wrote gets replaced? Etc. Now all
contributors are publicly acknowledged in the aforementioned repo,
and actual authorship is maintained by Github in a much more
effective and elegant way!
Finally, a few references to "Mycroft AI" were changed to the correct
legal entity name "Mycroft AI Inc."
==== Fixed Issues ====
#403 Update License.md and file headers to Apache 2.0
#400 Update LICENSE.md
==== Documentation Notes ====
Deprecated the ScheduledSkill and ScheduledCRUDSkill classes.
These capabilities have been superceded by the more flexible MycroftSkill
class methods schedule_event(), schedule_repeating_event(), update_event(),
and cancel_event().
* Add automatic uploading of wake words and learning option in enclosure client
* Spawn new thread and remove ping
Spawning a new thread eliminates the need to ping and eliminates additional latency. In addition, the return code of scp is now used to determine whether to delete the wake word
* Increment enclosure version number
==== Tech Notes ====
- Number of callers is now handled in the recognizer loop to make sure
status isn't lost when new settings is applied.
- Adds a force_unmute method to make sure the device is completely
unmuted when the user presses the top button.
==== Tech Notes ====
Autoloading configuration in submodules is bad for testing purposes and
should be reduced. It takes time and adds the possibility of altering
the base conditions for the tests
- mycroft/skills/core: global configuration moved to main.py
- mycroft/messagebus/client/ws.py global config moved to __init__ of
Websocket
- client/speech/mic.py global config moved to ResponsiveRecognizer
__init__() method
- client/enclosure/display_manager.py
- get_ipc_directory() called in methods where used
==== Tech Notes ====
Mute was previously a simple boolean meaning that if two parts of the
code wanted to mute it it would unmute the mic as soon as the first of
the two pieces of code wanted to unmute.
For example tts output will mute the mic and unmute it when the audio
output is complete. During first start up of a mark-1 the enclosure
mutes the mic and expects it to be muted until the pairing is complete.
This doesn't work. The mic get's unmuted as soon as the first sentence
has been spoken.
This counts the number of times mute is called and will not unmute until
as many unmute calls has been made.
Bug fix proposed for success wake-word recording.
Opt: Suggestion to change the save location of recording files as well.
Otherwise, the user could potentially get the permission denial.
-NeonGecko.com Inc
* BUGFIX: The big bug was calling is_paired() during wake_word_in_audio(). When not paired, that call hit the server, taking about a second. Since it happened multiple times a second, the audio buffers got backed up hugely. This resulted in weird behavior later as the buffers get cleared out.
* Added mycroft.api.has_been_paired(), which just looks for the pairing key (it does not validate it is still active with the server, like is_paired())
* The enclosure now checks for internet connectivity and kicks off the wifisetup process, not the wifisetup client itself.
* During the "onboarding" process, the microphone is muted using the new "mycroft.mic.mute" message. After pairing completes, the "mycroft.mic.unmute" is expected to be sent from the pairing skill. Unmuting again after a re-pairing is harmless.
* mute_and_speak() is smart enough to not unmute itself when complete if muted before
* util.check_for_signal() now accepts -1 as the lifetime. This means it never times out.
* util.stop_speaking() is more intelligent about shutting down the spoken text (including text that has been split at periods) and visemes
* Added a mycroft.api.is_paired() method
* Added mycroft.util.is_speaking and mycroft.util.wait_while_speaking() methods
* RESET now waits for the spoken notice to complete
* Stopped the "Checking for updates" and "Skills updated" prompts (commented out for now, probably will eliminate)
* Wifi setup filters out hidden ("x00") networks
* Visemes should keep up better if they get behind (will skip)
* Mimic is now searched for on the users path
* Onboarding process:
- wifi setup starts automatically
- User is walked through the process
- wake word and button pressing are ignored
- At end, a short tutorial is given
Improved the way the button press is handled. Now the listener gives the rest
of the system a moment to consume the button press before it interprets it
as a request to begin listening.
Also replaced the use of the "buttonPress" signal when "expect_response"
was indicated with an utterance. Now the signal "startListening" is
created, differentiating it from a buttonPress.
NOTE: this doesn't check for disk space and doesn't clean up
file stored in /tmp/ called mycroft_wake_sucess<time>.wav
TODO: some cleanup should be considered.
Signed-off-by: Art McGee <amcgee7@jaguarlandrover.com>
* The wake up sound is now played synchronously, thus not included in the recorded audio
* The minimum recorded phrase must be at least 0.5 seconds (instead of 0.1), and must be continuously quiet for that duration
* The silence threshold is reset every time we begin listening for a wakeword
* The silence threshold adjusts upward to just above the ambient sound while waiting for the wake-word
* Reformatted some comments to use Google-style docstrings, and added more comments
Main CLI enhancements:
* Microphone meter
* Long log line left/right scrolling
* Eliminated flicker
* VT100 ESC key code support (used by some terms)
In addition, to achieve the meter it was necessary to implement a mechanism for local Inter Process Communication (IPC). This is achieved using the file-system. By default a folder structure is created under /tmp/mycroft/ipc, but it can be directed to somewhere else by setting the config value in mycroft.conf:
"ipc_path" : "/path/to/somewhere"
In the future, Mark 1 and Picroft will get RAM disks to avoid burning out the SD card. This is also a very fast communication mechanism. This is all hidden under util.get_ipc_directory()
Further, the named signal mechanism was changed to use the IPC folder. The signal can have a lifetime now (not just one shot).
Max recording time is now 10 seconds instead of 30. This deals with cases where a noisy background prevents the listener's silence detection from triggering. 30 seconds was WAAY too long to keep listening -- nobody is going to be saying something that long for now.
* Several changes related to button pressing on the Mycroft unit:
- Pressing the button when it isn't listening starts it listening
- Pressing the button when listening will stop the listen
- Added a mycroft.util.signal() mechanism for out-of-thread communication
- Pressing the button now creates an "buttonPress" signal from the Enclosure
- The viseme playback and aplay check for the 'buttonPress' signal to abort
- Removed "Sorry I didn't catch that", irritating during false activations
* Fixed spacing that pep8 yelled about
* Added new listener config options
* Fixed audio unit tests
This adds a cleareraudio file for the wakeup test and changesto the new LocalRecognizer constructor
* Added .dict files to .gitignore
This is because they are now auto-generated on startup rather than stored permanently
* Fixed audio accuracy test for new LocalRecognizer constructor
* Added support for spaces in wake word config
In the phonemes a new word is indicated by a period character. The separating of the words actually changes the way pocketsphinx interprets the sound of it and in this case improves it
* Fixed unit test
previously the listner's behavior would be influenced by the buffer size. Now, however, it is not since the noise is increased by a factor of the seconds per buffer.
The 1980s birthed a new form of interaction between computers and users. For the first time computers became capable of understanding the most basic form of human communication - pointing and grunting. The mouse and the GUI revolutionized computing and made computers accessible to the masses.
We have now entered a third era. We are rapidly approaching a time when computer systems will understand human language and respond using the most natural form of human communication – speech.
This is an important development. Some might even call it revolutionary.
Despite its importance, however, the technologies that will underpin this new method of interaction are the property of major tech firms who don't necessarily have the public's best interests at heart.
Not anymore.
Meet Mycroft – the worlds first open source natural language platform. Mycroft understands human language and responds with speech. It is being designed to run on anything from a phone to an automobile and will change the way we interact with open source technologies in profound ways.
Our goal here at Mycroft is to improve this technology to the point that when you interact with the software it is impossible to tell if you are talking to a human or a machine.
This initial release of the Mycroft software represents a significant effort by the Mycroft community to give the open source world access to this important technology. We are all hoping that the software will be useful to the public and will help to usher in a new era of human machine interaction.
Our community welcomes everyone to use Mycroft, improve the software and contribute back to the project. With your help and support we can truly make Mycroft an AI for everyone.
Joshua W Montgomery – May 17, 2016
Matthew D. Scholefield
Åke Forslund
penrods
Matthew D. Scholefield
Jarbas
jarbasai
reginaneon
kfezer
kfezer@gmail.com
Art McGee
SoloVeniaASaludar
Jonathan D'Orleans
Ethan Ward
Arron Atchison
Ryan Sipes
Leo Arias