Simple app to deploy on Heroku to test two endpoints. This is part of a project during Lambda School.
https://github.com/Build-week-picmetric2
My practice Heroku instance, might use for production
- Student contributed idiomatic and stylistically correct code to the project.
Idiomatic: tested with PEP8
- Problem they had to accomplish on their own:
To proceed with testing of API vs. Michelangelo's layer, I created a Heroku instance, and successfully tested with Michelangelo.
Show code (above)
Break down their participation in the project.
-
Created and tested (on Heroku) endpoints that interacted correctly with Michelangelo's work.
-
Communicated with Chris, who developed similar code on AWS.
-
Identified critical need for desctription of return value from classifier.
- Student used and was able to implement and talk technically about their understanding over the database/data-pipeline used for the project. They also may have participated in helping design the schema.
I discussed with Michelangelo and team storage alternatives (Bucket, NoSQL, SQL) for different types of data.
- Images received from user - bucket
- User, transaction data - Postgres
- Results from classifier - Postgres, string field
I participated in designing model, by proposing storage of pre-processed image, by discussing which data is needed by the endpoint layer, and other points.
- Student participated in and contributed to the routes/API relevant for Data Science functionality.
I did this by implementing the required endpoints in a python script hosted on Heroku, Functionality was tested by Michelangelo.
-
Ensure here that student can speak technically to the frameworks/languages/tools used to create the API and ensure that the data flowed smoothly across the project back-end.
-
Framework - Flask was a natural choice because it is easy to use, and ue of it and Heroku are a known quantity. For testing, this accepts GET requests. In production, only POST requests will be needed.
-
Languages - I only used Python (and JSON).
-
Tools - I used Heroku as infratructure. I attempted to get an AWS account, but could not get my card validated. I am pursuing this.
-
Smooth flow of data - as of Tuesday afternoon, Michelangelo was happy with data flow between his servers and the endpoints. Connection to the model built by DS9 is incomplete, but identified as a requirement.
- demonstrates that all MVP features were built
Two endpoints reqired (/summary and /batch_img_summary) are defined and worked as of Tuesday afternoon.
Features:
-
/summaryrecieves image URL in JSON via POST request, returns JSON containing classification. -
/batch_img_summaryreceives JSON with multiple copies of data as for/summary, and returns JSOn containing multiples of return for'/summary
- Student successfully collaborated with colleagues, "translated" DS topics for non-DS peers, and handled any problems or friction appropriately.
I think so. I cooperated successfully with Michelangelo and Chris, and got a commitment to a data format to be returned from the model from Marilyn and Todd. This allowed Michelangelo to continue with work on storage of result instances.
I had some friction with one team member. Not a show stopper (yet), but cooperation regarding updates to git repo could be smoother.
Team Lead: Nicholas Gallucci
My Awesome DS TL - Jordan Ireland (Leader of Sharks)
DS (Unit 3) Chris Huskey (Laser Shark)
Michelangelo Markus
DS (Unit 4) Marilyn Esko
DS (Unit 4) Todd Gonzalez
Natalya Beckstead - User Interface
Eian Carter - Front End
Kevin Jensen - Front End