Mr Akountant

Why download an app to track expenses when you can have your own personal accountant with you 24x7?

Comment

Inspiration

There are so many smartphone apps that track personal expenses. But pulling out a phone, typing out the expense, selecting a category...too slow! Just when you walk into your house/office space where there is a Google Home Mini or when your phone's google assistant is listening, speak out your expense and Mr Akountant will do the heavy lifting.

What it does

Mr Akountant is a Google Assistant app that listens to the expenses you did and keeps a track of them. It supports categories such as movies, gas, internet, shopping etc.

How we built it

  • We have designed various dialogues that can happen between the user and Mr Akountant in Google Cloud Platform's Dialogflow service. Dialogflow intents (conversation flows) are then translated as Google Actions which further act as actions inside the assistant application.
  • Dialogflow after the parsing the input from the user, sends it across to a AWS Lambda function which is where the business logic resides.

Challenges we ran into

  • We had never used Dialogflow, Actions and Google Assistant. Major challenge was to understand how natural language processing works in Dialogflow and design the AWS business logic to support the output of the NLP.
  • Both of the team members do not primarily work on python, but to get a learning experience from this hackathon, the AWS lambda functions have been built in Python. Stackoverflow to the rescue!

Accomplishments that we're proud of

  • We have achieved the full integration from the assistant application to the business logic in AWS.
  • All the boilerplate is now ready and only time it will take is to add features. Database, APIs, dialogs are all ready. Think of new scenario? Add the dialog, write the API and done! Assistant now supports that new scenario.
  • Never worked with the GCP service, still proud to have achieved the end to end integration.

What we learned

  • Think extensively about possibly infinite ways the user can give a voice input to the assistant application.
  • Getting requests that travel from GCP to AWS work. Never thought, AWS security groups would be such a pain.
  • How different accents can trigger different NLP scenarios and tackle those.

What's next for Mr Akountant

  • Integrate with splitwise APIs to enable maintaining a record of how much you owe to your friends and how much you are owed.
  • Integrate with PayPal APIs to pay to your friends (when you owe money) or request your friends for money (when you are owed).

Built With

Share this project:

Updates