- `stop_conversation_on_speech_match` should default to True.
- `render_response` should also handle conversation follow-ups, set the
follow-up to True if the response ends with a question mark and the
value of `with_follow_on_turn` is not set,
- Don't render responses if a `tts_plugin` is not set.
YAML isn't part of the Python standard library, while JSON is.
If we want `setup.py` to dynamically parse the available integration
manifest files in order to populate the extra dependencies, then it's
better to rely on a JSON format for manifest files - the parser is part
of the standard library and it doesn't require the user to install
`pyyaml` before `platypush`.
The assistant object now runs in its own thread and leverages an
external `SpeechProcessor` that uses two threads to scan for both
intents and speech in parallel on audio frames.
- Added `intent_model_path` parameter.
- Always apply `expanduser` to configuration paths.
- Better logic to infer the fallback model path.
- The Picovoice Leonardo object should always be removed after
`assistant.picovoice.transcribe` is called.
The plugin should leverage `AssistantPlugin._on_mute_changed` to handle
the boilerplate state managent on mute/unmute actions instead of
re-implementing the same logic.
- The `Responding` state should be modelled as an extra event/binary
flag, not as an assistant state. The assistant may be listening for
hotwords even while the `tts` plugin is responding, and we don't want
the two states to interfere with each either - neither to build a more
complex state machine that also needs to take concurrent states into
account.
- Stop any responses being rendered upon the `tts` plugin when a new
hotword audio is detected. If e.g. I say "Ok Google", I should always
be able to trigger the assistant and stop any concurrent audio
process.
- `SpeechRecognizedEvent` should be emitted even if `cheetah`'s latest
audio frame results weren't marked as final, and the speech detection
window timed out. Cheetah's `is_final` detection seems to be quite
buggy sometimes, and it may not properly detect the end of utterances,
especially with non-native accents. The workaround is to flush out
whatever text is available (if at least some speech was detected) into
a `SpeechRecognizedEvent` upon timeout.
- Added wiring between `assistant.picovoice` and `tts.picovoice`.
- Added `RESPONDING` status to the assistant.
- Added ability to override the default speech model upon
`start_conversation`.
- Better handling of conversation timeouts.
- Cache Cheetah objects in a `model -> object` map - at least the
default model should be pre-loaded, since model loading at runtime
seems to take a while, and that could impact the ability to detect the
speech in the first seconds after a hotword is detected.
`AssistantEvent.assistant` is now modelled as an opaque object that
behaves the following way:
- The underlying plugin name is saved under `event.args['_assistant']`.
- `event.assistant` is a property that returns the assistant instance
via `get_plugin`.
- `event.assistant` is reported as a string (plugin qualified name) upon
event dump.
This allows event hooks to easily use `event.assistant` to interact with
the underlying assistant and easily modify the conversation flow, while
event hook conditions can still be easily modelled as equality
operations between strings.
The integration was based on my old fork of the AVS service, which is no
longer functional given the changes the the Amazon's backend side.
A new `avs-device-sdk` is now available, but it seems that it requires
lengthy compilation processes which are RaspberryPi-specific.
Further investigation is needed for a new Alexa plugin - see #334.
It only existed as a back-compatibility layer with armv6, since there
was no build of the assistant library that worked on Raspberry Pi Zero.
But that API layer has been discontinued by Google and it's no longer
functional, so only the `assistant.google` integration (on x86_64 and
armv7) is currently supported.
Added an `add_dependencies` plugin to the Sphinx build process that
parses the manifest files of the scanned backends and plugins and
automatically generates the documentation for the required dependencies
and triggered events.
This means that those dependencies are no longer required to be listed
in the docstring of the class itself.
Also in this commit:
- Black/LINT for some integrations that hadn't been touched in a long
time.
- Deleted some leftovers from previous refactors (deprecated
`backend.mqtt`, `backend.zwave.mqtt`, `backend.http.request.rss`).
- Deleted deprecated `inotify` backend - replaced by `file.monitor` (see
#289).
- If a Python optional dependency is available as a system package on
the target system, try and install it that route rather than pip. It's
usually faster and it decreases the risk of breaking system packages.
- Added support for apk dependencies in manifest files. This brings the
number of distros officially supported by all the extensions to four:
- Alpine
- Arch
- Debian
- Ubuntu
It was just too painful to find a combination of versions of gunicorn,
gevent, eventlet, pyuwsgi etc. that could work on all of my systems.
On the other hand, Tornado works out of the box with no headaches.
Also in this commit:
- Updated a bunch of outdated/required integration dependencies.
- Black'd and LINTed a couple of old plugins.
