Jarbas and machine learning results

[cuducos]

I'm documenting here the next steps of Jarbas in terms of development, i.e. it's a big issue, not a tiny and specific one.

We have a new player in Serenata de Amor operation that is Rosie. She's our robot, she'll tell us what she thinks of each reimbursement. Basically for each hypothesis of our roadmap Rosie will:

1. Say that for each reimbursement uniquely identified by a certain combination of year, applicant_id anddocument_id there is the a probability of irregularity/immorality/illegality (i.e. 0 < probability < 1) .
2. Send a HTTP POST request to Jarbas with this probability and an some meta information of the hypothesis this probability refers to.

Thus Jarbas should be able to: (removed in favor of this list)

Conceptually this means that Jarbas is moving from an internal tool to help people engaged in Serenata de Amor to a public data visualization platform to a share our artificial intelligence machine learning results with the world. A lot of changes (refactor, renaming, endpoints etc.) are expected 🎉

[cuducos]

This just started at next branch ; )

[cuducos]

I just reslized the TODO list from the previous post changed considerably, so I'm gonna edit the post, remove the previous TODO list and use this one ro keep track of the progress:

• Create a new Reimbursement model to fit the new YYYY-MM-DD-reimbursements.xz
• Add fields to that model to store data about suspicions Rosie might find
• Add the current Receipt logic to Reimbursement
• Create list and detail API views for Reimbursement
• Create list API view for /api/subquota/
• Create list API view for /api/applicant/
• Use filter parameters in all three list API views
• Create command to read YYYY-MM-DD-reimbursements.xz
• Create command to read YYYY-MM-DD-irregularities.xz (generated by Rosie)
• Use cache in API views
• Adapt the front-end to new data structure (moved to New UI/UX #61)

[cuducos]

Hi @daneoshiga: The changes described in that Issue will break compatibility with tapioca-jarbas (check the README.md from the next branch for more details). So I'm tagging you here in case you want to coordinate an update to your repo — we can make it happen ; )

[daneoshiga]

Thanks for the heads up, I'll keep keep an eye in the evolution of it and update tapioca-jarbas accordingly, thanks :)

[cuducos]

@daneoshiga soon I'm merging next to master, so the new API will be available. However until we write the new front-end #61 the old API still working because our current front-end consumes it. No need to hurry, but anyway: if you wanna pair I can learn bits of tapioca and maybe we both can solve it quicker ; )

[cuducos]

Closed by #64

[daneoshiga]

@cuducos sorry for taking too long to reply.
We could do that :), I have free time at night these days, but tapioca-wrapper is pretty easy to work with, just listing the new endpoints in resource_mapping.py should work :)