Improving the Alexa Skill Bus Salamanca

In this post, I share the improvements made to the Alexa skill "Bus Salamanca", as well as some lessons learned and challenges encountered during the process.

Deployment Automation

• I added a GitHub Workflow to automatically deploy to AWS Lambda, including credential management via IAM.
• There are now two Lambdas: one for dev and another for prod.
• AWS errors have been very descriptive when roles didn't work, which allowed me to fix them quickly.

Environment Management

• I learned that having two environments (dev and prod) in the skill is straightforward.
• To publish the development skill, you need to specify the production Lambda.

Improvements to the Graphical Interface (Alexa APL)

• I used Alexa APL to enhance the skill's graphical interface.
• The interface adapts to different devices, such as the Echo Spot (small circular screen) and larger devices like the Echo Show.
• Now, useful information is displayed clearly, requiring scrolling only when necessary.

• I used the Alexa Developer Console to design the graphical interface and ensure it adapts perfectly to different devices.

• I also improved the initial screen when there is no data, prompting for the stop and showing voice commands to query any stop.

• This is how the upcoming bus lines are displayed:

Technical Considerations

• I discovered that the Lambda runs in a different time zone, so you need to adapt to that.
• After getting the first approval for the skill, the second one took barely an hour to be approved.

User Experience

• I spent more time testing the skill to add better recommendations on devices like the Echo Show, such as suggesting to check the bus once the stop is saved.
• I improved the initial state the first time the skill is shown, making it more intuitive to request information.

Optimization and Robustness

• I tried to optimize the skill, but there are two main points that are hard to improve: contacting the external service and transforming XML to JSON.
• I improved the information shown when the service does not respond or there is no data, preventing the skill from hanging and always providing a response to the user.

Environment Differentiation

• I added a simple way to know if the skill is dev or prod, by adding an emoji to the title of the cards shown on screen.

Final Thoughts

I'm considering publishing the code on GitHub, but I'm not sure if it's legal, since the Salamanca API is not public and I don't know if I can share it.

If you want to try the Alexa skill, you can find it here: Bus Salamanca on Amazon Alexa