velicanu / expenses Goto Github PK

Steps:

1. Delete Files
2. Pick a file
3. Check confirm delete
4. Pick another file

One of the pain points of using this app is that it's difficult for non-programmers to set it up.

We can try making a "website" run this so a User can log in and see their expenses.

Required to make this work:

• some form of username / password authentication. Can start by doing this through streamlit.
• some form of data sandboxing per user.

Optional add on:

• encrypt each user's data using their password, only decrypt them in RAM when using the app. at rest they should be encrypted and un-readable by the server operator.

Right now when a credit card transaction doesn't have a category, it defaults to Other. The current workflow to properly characterize "Other" transactions is to edit the python code to add a rule for it.

It would be nice to be able to add/edit/remove custom rules completely from the UI and auto update the categories as the rules are tweaked.

add an end-to-end test of the full pipeline, using the sample data

Right now only 1 term can be searched at a time by using the description box.

Support the following usecases:

uber+lyft --> search for uber OR lyft
uber*eats --> search for uber (wildcard) eats
uber^!eats --> search for uber AND NOT eats

in addition to cc transactions, maybe we want to include manual transactions, such as rent / venmo / etc.

add this as a new data source.

cleanup bash scripts

Given an input like:

!wayfair^!REI^!Google^!Micro^!Best^!START^!OONI^!THURO^!SQUAW^!SAIKO^!Amazon^!Hard^!Whole^!Home^!ebay^!sitonit

Show a pie chart of each category along with everything else

right now the test phase of github actions is slow mainly due to running pip install -r requirements.txt each time, there's probably a better way to do this using either caches or a docker image.

potential idea: https://medium.com/ai2-blog/python-caching-in-github-actions-e9452698e98d

to improve user experience we'll auto-detect which card type a data file is from the columns.

dropping the filename convention

Right now, source appears as amex for both Amex Platinum and Amex Blue Cash. It would be better if these two appeared separately

It would be nice to be able to fix errors / update data directly in the UI. This looks like a promising way to do this on the frontend: https://github.com/PablocFonseca/streamlit-aggrid , need to tie it in to the back end as well so that changes persist

Need to take a step back and look at the frontend code to see if it can be organized better. a lot of features have been added and the code in src/streamlit_app.py has become a little challenging to work with

User logic got removed since it broke from updating streamlit. Find a new solution to do auth and login.

something about tempfile is crashing on windows

Right now we have to manually download, then upload, csvs from every card whenever we want to see updated expense data which is a pain.

It would be really nice if we could click a button and have the app automatically pull the latest data for each card we've added.

The Plaid API can make this happen.

Right now clicking the Run Pipeline button in the UI throws an error if no data has been uploaded, should instead say something more useful.

Also add some instructions / guide how to get started for new users.

Right now whenever the app is opened from scratch, all the date ranges and selections reset to default values.

It would be better if they defaulted to the last used values, for example the date range since usually the most interesting data is recent, not since the beginning.

barclays looks like this

Transaction Date,Description,Category,Amount
12/27/2019,"Payment Received","CREDIT",1469.31
12/25/2019,"MYNTRA DESIGNS PRIVATE","DEBIT",-21.90
12/22/2019,"UBER TRIP HELP.UBER.COM","DEBIT",-8.09
12/21/2019,"PEETS COFFEE","DEBIT",-17.09
12/20/2019,"DUNKIN #356593","DEBIT",-2.39
12/19/2019,"CHULO RESTAURANT & BAR","DEBIT",-62.39
12/17/2019,"TST* AREA FOUR RESTAURANT","DEBIT",-3.68
12/17/2019,"STARBUCKS STORE 08742","DEBIT",-8.45