openkfw / qupo Goto Github PK

In deployment the env variables in the api service are not set. They are set via a config file taking from the .env file. In deployment the .env file doesn't exist, so they have to be taken directly from the environment. Restructure the api service to take the variables directly from the environment and not require an env file.

When sending the command to calculate the portfolio with the ionq algorithm, an error in the backend appears.

• Handle errors in the backend properly with error messages.
• Show in frontend if the calculation didn't go throught (e.g. with snackbar) and handle the ui, so it wouldn't result into breaking the app
• Sometimes symbols are not found and therefore not processed. If in a calculation request there are no or only one symbol found, the calculation can not be done. A proper error message and feedback for the user which symbols didn't work should be provided.

• Integrate pyportfolio for calculation of financial KPIs
• Integrate pyportfolio solver suite into backend code

• Add sqlite as dependency to requirements file of api
• Connect local sqlite file to backend code
• Add possibility for user to choose whether a sqlite databse should be used at all (e.g. CLI parameter)

Currently only basic filtering to search through indices, countries or industries is in place. Extend filtering capabilities to search for single symbols and company names on each view. Show all indices, countries, industries filtered accordingly that contain the symbol or the company name fow which is searched for.

• filter with symbol.name
• filter with symbol.symbol
• when search by symbol - set extended view true

• When adding symbols (<StocksCollection /> component) only searching for symbols is possible. Include also the capability to search via company names for the correct symbols.
• Show company names in chart instead of the symbols

Add some documentation to the repo. Check if the charts of the db structure are up-to-date and show proper architecture charts with an explanation in an .md file. If necessary make some refactoring changes and/or add some comments, so that the code base is easier to understand.

• Find reliable source for collection of stock ticker symbols (e.g. DAX)
• Generate a list of valid stock ticker symbols based on collection
• Use list as input for #2

• Show more information about the companies
• Explain the Upright Project of the net impact ratio
• Create an info page

Once calculated the results, the parameters should be able to be adjusted via the UI. A button should be provided to recalculate the results with the changed parameters (risk_weight, esg_weight).

This validates if the existing data works fine

A lot of symbols that are send to the api are not being found by the yahoo library. Maybe exchange the library of the symbols to provide working symbols or find another solution to prevent sending symbols to the api that can't be processed.

• Fetch data from yahoo finance for given time frame (e.g. 01.01.2018 - 31.12.2018)
• If "save to DB" is enabled (see #1)
  • Check if data for given stock in given time frame is available in the DB
  • If yes, get data from DB
  • If no, get data from yahoo finance online and save data to DB
  • If partial data is available, fetch all data from yahoo finance and update DB entry

• Identify lists of stocks that can be used as example inputs/problems
• Generate small workflow where the user can choose different indices and number of stocks and a list of stocks for data fetching is automatically generated by #3

Once a request is sent to the api, the api is busy and doesn't take any further requests anymore. In deployment this is a problem. The health check in the helm charts for the api is currently disabled due to this.

Rebuild the api to take several requests and then change the helm chart to allow the health check on the api /health endpoint.

Following algorithms still have to be connected to the frontend, to show in the chart:

• PyPo
• Qio
• Qiskit
• Ionq

Follow-up of Pull Request 21

Idea: Make different setups (stock symbols, time frame, input parameters) identifiable and save the job result to the database. If the same setup is requested at a later point, the calculation does not have to be run again, but rather just read from the DB.

View names of the companies in addition to the symbols for a better user experience.

It is not clear which model is doing what and where they are coming from (for example the qiskit algorithm is from IBM). The UI could show some information on that.

use GDELT to get the current news sentiment on a given company from the current portfolio.

On the /portfolio page when moving the sliders and thereby changing the values of esg_weight and risk_weight the classical models should be recalculated on change and not on button calculate click. So the results shown in the chart can be changed faster.

When starting the application with the setting use_db=False, the sqlite db is not used. In some functionalities this leads to bugs, e.g. when sending a request for calculating the osqp algorithm.

The history of the stocks are currently fetched only for a specific timeframe, make it editable to the user.

When running npm install for the frontend, a Could not resolve dependency error occurs. Solve dependency conflicts.

Should be possible to calculate the portfolio also without quantum?

On the portfolio page (or maybe also somewhere else) show already calculated results in a list sorted by entry created, so that the latest calculation is the first entry in the list. Following details should be shown in the list:

• the stocks that were included in the calculation
• the timeframe that was chosen
• the weights (risk, esg)
• the calculated models

streamline config and setup of project environment with poetry setup (similar to npm)

compare result quality and efficiency

• PyPO vs OSQP
• classical vs quantum inspired (Azure)
• classical vs QC (IonQ)

Instead of only being able to choose from a set of symbols depending on the view, create a component to allow a selection of single symbols and then Continue with these symbols to calculate the models.

The behavior of the quandl library is not as expected. Make sure the call to the nasdaq api returns a big enough dataset to extract the net_impact_ratio. Maybe look for another python package to make the call.

Implement a way to get back to the tickers view (http://localhost:3000/). Maybe by having a button Restart process during the process and once on the portfolio page.

• setup runscript from PO model dataframe to optimization result
• OSQP
• QIO
• qiskit ionq backend via azure quantum

Restructure and rename: following the Clean Code guidelines. Look through comments and To-Dos and solve all issues regarding refactoring.