Skip to content

CHANGELOG.md

Latest commit

 

History

History
64 lines (45 loc) · 5.21 KB

CHANGELOG.md

File metadata and controls

64 lines (45 loc) · 5.21 KB
Raw

Change Log

v2.0.0

MAX Toolkit v2.0.0 is a major release that now allows Agents to communicate with MAX via Inter-Process Communication (IPC). This is facilitated by the MAX IPC framework which leverages gRPC C++ to provide IPC capabilities.

MAX-API Changes

  • Added a new HashableInterface which is used by the ActivityRequestInterface and the DialogRequestInterface (required for use cases introduced with IPC).
  • Methods in StartListeningCallback and ActivityRequestInterface are made virtual.
  • A new registerAgent() API is introduced in the AgentManagerInterface, to support single agent registration use cases like agent process crash recovery.

Enhancements

  • The MAX IPC framework is introduced along with two major packages - MultiAgentExperience-IPC and MultiAgentExperience-IPC-gRPC. The MAX IPC framework facilitates inter-process communication between the MAX process and any agent processes, and enhances security by isolating processes.
  • Added a new package - MultiAgentExperience-ThreadExecutor.
  • The MAX SampleApplication is updated to incorporate the MAX IPC framework. Read more here - MAX_IPC.md.

Issue fixes

  • Added a fix to address an issue which did not allow Alexa alerts to be stoppable after a power-cycling the device.

Known Issues

v1.3.0

MAX-API Changes

APIs supported by MAX Library are provided in a separate MAX-API Directory with this release. The details of what has changed in the APIs with respect to v0.4.0 can be found in the below document. MAX_API_Documentation.md

Enhancements

  • Added the ability to temporarily store any activity/dialog requests in a buffer, during agent re-registration, when the MAX process recovers from a crash. After all agents have executed their OnCompletionCallback (received in AgentRegistrationInterface::onReady), the requests stored in the buffer will be processed in a priority order based on the activity type of the request. This helps restore correct state of activities/dialogs when after a crash.
  • Changed the set of shared ControlTypes given to an agent during a dialog from “any controls available for other actors” to “only the controls for foreground experiences of other actors”.
  • The logger has been extracted out of MAX and bundled separately in this package named MultiAgentExperience-Logger. This doesn't disturb any client operation but helps make the code more modular.
  • Added MAX integration for pending notifications.
  • Added MAX integration for 'Do Not Disturb'.
  • Added integration for UDC in Alexa Cloud and AVS-SDK.

Issue fixes

  • Added a fix in AVS-SDK to address a race condition causing Alexa alerts to not stop on invoking a UDC STOP command.

Known Issues

All issues listed here are the behaviour/ issues observed with one agent as Alexa.

  • Alexa alert is not stoppable using UDC after power cycle - If a user restarts the device while Alexa is ringing alert, after the device boot up, Alexa alert is not stoppable using UDC.
  • Content(News) mixability type is set as mixable-restricted instead of non-mixable in some cases - If a user plays music from Alexa app, then the user requests for news from Alexa followed by a dialog from the other agent, then news gets attenuated while the dialog is ongoing. News is supposed to be paused instead. This happens when “Amazon Music” or “Tune-in Radio” is played from Alexa app.
  • Content(Music) mixability type set as non-mixable instead of mixable restricted in some cases - If a user plays news first and then music(“Amazon Music” or “Tune-in Radio”) from Alexa app, then music gets paused insted of getting attenuated.
  • Reminder's dialog/TTS is not delivered when reminder is enqueued with timer - A reminder has an audio cue followed by TTS indicating a reminder message. If user sets the timer first and then a reminder, timer starts ringing first and reminder is enqueued. When the user stops the timer, reminder audio cue is played but the TTS is not delivered.
  • Music does not attenuate when notification is received in some cases- If a user plays music (“Amazon Music” or “Tune-in Radio”) from Alexa app, and then if a notification is received, music does not get attenuated and plays in full volume. The music is supposed to get attenuated in this case.

v0.4.0

Initial release of the MAX Toolkit.

Known Issues

Sample Application

  • When running the SampleApplication with wakeword engines, you will see false positive errors in the console regarding wakeword engine creation. These errors can be safely ignored.
  • The MAX Sample Application will become unresponsive if you attempt to “barge in” on an un-interruptible response (a response that contains either wake word “Alexa” or “Computer”) from Alexa. For example,
    • Say “Alexa, what is Alexa Voice Service?”
    • Then while she is speaking “Alexa, stop”.
  • The Alexa agent may lose context in multi-turn interactions. Alexa will respond with “Sorry, I don’t know that” (or something similar) to your response.
  • Alexa will sometimes become unresponsive after multiple music or content streaming activities.
  • Ubuntu: The Sample Application will sometimes segmentation fault during shut down.