My recommendation would be to start with the Hello World module and implement a feature at a time.

Get it to query the API and display the results in your module. Get it to respond to a message sent from any another module. Get it to respond to a specific message with a structured payload. Configure the voice recognition module to send the specific message & payload.

There’s a pretty good set of instructions in the Developer section of the Magic Mirror Github repository.