Metadata
Preferences
Misc User Info
Conversation History
Speech to Text
User speech:
Let’s go for a walk
Location Agent
Nearby landmarks
Planner Agent
Decide what to do next
Data Agent
Retrieve live wikipedia, google reviews
Travel Agent
Recommend based on metadata, update metadata based on conversation
Coordinator Agent
Decide what to say
Text to Speech
Agent speech:
How about … ?
Retreive
Update
metadata
Button clicks are treated as user speech