diveflo / fair-teams-ai Goto Github PK

GitHub changed the way dependabot pull requests are handeled.
TL;DR: Dependabot PRs cannot access secrets defined for the GitHub actions, which makes our tests fail currently.

We should remove skill, profilePublic and steamName from the request body because it is not needed. As a first step this attributes should be optional, so the application is still working after the api change.

[
  {
    "name": "string",
    "steamName": "string",   //remove
    "profilePublic": true,   //remove
    "steamID": "string",
    "skill": {}   //remove
  }
]

Add optional queryParameter hltv=true/false to scramble teams by hltv

Eieiei wer macht denn sowas ¯_(ツ)_/¯

We have a hard-coded Steam Web API key in the Player.cs class....we should discuss how to correctly handle such secrets without adding too much extra efforts during development.

api endpoint structure

request:
{
    "players": [
        "name1",
        "name2",
    ]
}

response:
{
    "statistics": [
        {
            "description": "",
            "values": [
                10,
                20,
                30
            ]
        }
    ]
}

same order in response as in request. Or do we need no request body and deliver the stats of each default player?

Currently we have a hardcoded list of players in the webui, with a hardcoded map to their SteamID in the frontend.

It would be great if we could add players to that list during runtime/in the webui if someone new wants to play with us spontaneously.

I'll add required tasks to enable this for both the back and frontend.

Dude. I didn't see any changes when using the map selector...Niggo noticed and told me that it actually does something :D

Currently, unit testing feels clunky as the different ratings (HLTVRating, KDRating, DummyRating) are created within SkillBasedAssigner rather than somehow provided from the outside via DI or something similar.

Rethink and (as appropriate) rework design.

Currently, a wrong or missing SteamID results in an exception from the backend as it can neither find the corresponding steam user name nor (once that's implemented) scrape the rating from csgostats.gg.

If/once we want to add the possibility to add new players live/in production on the frontend page, the backend needs to be able to gracefully handle such a case:

• use the handed over name instead of the retrieved one from the Steam API
• use some default rating for the player

As we still will want to notify the user that the SteamID wasn't valid and therefore a default rating was used, we need to figure out how to gracefully fail while still returning the expected data in ASP.NET Core
Maybe helpful documentation

Currently the frontend shows just empty whitespace in case we don't select an even number of players.

To clarify the output, we should add "BOT" or something similar instead to get an even number of players.

As we now have separated workflows for the frontend and backend, we should configure our CIs to only run the corresponding workflows if changes have been made to that codebase, i.e., only run the backend CI when something in the backend changed and vice versa.

Currently we use sqlite as our db provider for the backend.
As this is very entangled with the backend itself, it causes some downtime for the backend even if we just need to make some fixes for the db.

Using a different db, e.g. PostgreSQL or MariaDB, as a separate docker container, we could reduce this downtime.

Currently, if the computed score of a player is "round" we display less digits than for the other players.
This triggers my OCD :D

Currently, our CI/GitHub actions manually calls flutter build web and subsequently uses the flutter-gh-pages action to deploy the resulting app.

The flutter-gh-pages action should however, at least if I read this file correctly, already run flutter build web. So it looks like we're building it twice at the moment.

Currently our frontend is deployed on GitHub pages while the backend is deployed to a linode VM.
The backend REST Api is wrapped in an nginx proxy hosted locally on my network.

Deploying the front and backend together would have multiple advantages:

• better security, as we wouldn't have to allow any outside-network requests in the backend
• higher availability as outages in my local network won't disturb the nginx proxy