For the past two years I have been using the MBTA Bus Tracking Skill and it has performed flawlessly but recently stopped working properly.
The skill requires three pieces of information - bus route, direction and stop - to announce bus arrival predictions and uses a conversational style to prompt for the information. When it is working properly the flow is:
User: t bus arrivals
Mycroft: name a route
User: one
Mycroft: choose a direction Outbound toward Harvard Square or Inbound toward Nubian Station
User: outbound
Mycroft: where will you be getting on?
User: 77 mass ave
Then Mycroft lists all bus predictions for the stop.
What just started happening is:
User: t bus arrivals
Mycroft: name a route
User: one
Mycroft: choose a direction Outbound toward Harvard Square or Inbound toward Nubian Station
User: outbound
Mycroft: I don’t know what that means.
I can see in the CLI the QuestionsAnswersSkill swooped in and tried to handle the response.
ReSpeaker_4mic_hat | Pixel Ring: Think
18:18:02.948 | INFO | 682 | ReSpeaker_4mic_hat | Pixel Ring: Off
18:18:02.977 | INFO | 682 | QuestionsAnswersSkill | Searching for outbound
18:18:02.984 | INFO | 682 | ReSpeaker_4mic_hat | Pixel Ring: Think
18:18:04.070 | INFO | 682 | QuestionsAnswersSkill | Timeout occured check responses
18:18:04.917 | INFO | 682 | ReSpeaker_4mic_hat | Pixel Ring: Off
What is weird is if instead of saying “inbound” or “outbound” I ask something that should be handled by the QuestionsAnswersSkill the Bus Tracking skill continues.
User: t bus arrivals
Mycroft: name a route
User: one
Mycroft: choose a direction Outbound toward Harvard Square or Inbound toward Nubian Station
User: how tall is john cleese
Mycroft: where will you be getting on?
The skill uses fuzzy match for direction. It looks like the direction is set correctly if the prompt is answered Jeopardy stile:
Mycroft: name a route
User: one
Mycroft: choose a direction Outbound toward Harvard Square or Inbound toward Nubian Station
User: what is inbound
Mycroft: where will you be getting on?
From there the skill finishes successfully.
As I said, it’s been working for two years. I’m wondering if something changed in the way conversational style prompting is handled.