[Meetings] Nov. 3 Meeting

Prior to Meeting

We created a rough sketch of the codebase the prior weekend, but there are still questions and concerns we should address:

• Packages have inter-dependencies (e.g. speech speed will require the speech-to-text package and the voice intensity package -- the latter would be used to determine what counts as a pause)
  • We could address this by defining a hierarchical relationship between packages or by having packages define classes and their instances could freely interact with each other (Taran prefers the latter)
• What defines how packages interact?
  • We could define higher-level "controller" classes that direct operations -- this might also remove the need for packages to interact with each other
• Do we want to have a traditional server/client implementation?
  • The alternative is hosting on something like Firebase and using cloud functions
  • Flask is customizable, but Django offers a lot on installation, like authentication and database migrations
  • Could we write the front-end using React?
    • We could also have the back-end offer an API and the front-end separately builds on Netlify -- this way, it'll be easier to build them in parallel and independently
• We'll also have to think about how we're doing version control for everything machine learning-related
• We'll have to discuss dev-prod parity (specifically, how things might change on a remote host) so it doesn't bite us later on
  • We can use Docker for a developer environment to match up to how production builds, and we'd all have the benefit of not encountering problems of people being unable to install certain packages
• Since this is a multi-month project, unit testing is crucial
  • The project already offers the ability to do this type of testing
  • Do we want any other types of testing?
• Taran would prefer everybody in the group develop in Linux, since he has experience deploying on the OS, and it's what servers typically use. The Docker image currently in the repo is based on Linux, so everybody is free to continue using Windows if they use it

Meeting Conclusion

• Packages define classes and instances may interact with each other as directed by controller classes
• Flask as a server
• No resolution reached regarding whether we're separating our front-end and back-end architectures (Taran is leaning toward separation, but will defer decision until we need to work on the GUI
• Team agrees to use Docker and support Linux as the production environment
• Majority feel that testing is unnecessary
• Additional work will need to be done to investigate how we're splitting up gesture recognition into smaller tasks