Skip to content

Intro: run the backend locally without the AWS account #6

Closed

Description

@alessiopagliarulo

Context: Deep Learning Playground (DLP) is a web app where someone uploads a dataset, picks settings, and builds a PyTorch neural network by dragging and dropping layers, no code needed. It was built by an earlier DS@GT team and has been dormant since late 2024. This semester (Fall 2026) is our team picking it back up. The first two weeks are an intro period: learn the basics, get the project running, and get to know one corner of the code. Nothing here is expected to be perfect, and asking questions in Slack is encouraged.

Status checked on 2026-10-13 (on a Mac): the website starts and its pages load, pnpm test and pnpm lint pass in frontend/, all 18 backend tests pass in training/, and the Django server starts with its API docs at /api/docs. The training screen redirects to login, and login needs a Firebase project the club does not have yet. Do not set up anything on AWS or Firebase; club leadership asked us to hold off until ownership is confirmed.

How to run things: see AGENTS.md (section "Run") and README.md in the repo root.

Task (intro, about 3-5 hours)

Get the backend (training/, Django + Celery + Redis) running locally without the AWS account, and write the steps down.

Why

The documented one-command setup (docker-compose up in training/) mounts ~/.aws and expects an AWS login from the old team (AWS_PROFILE=dlp). None of us have that. Without Docker, the server does start (tested 2026-10-13 using the conda + Poetry steps in AGENTS.md), but actual training jobs go through Celery and Redis, which that path does not start.

Steps

  1. Try the Docker setup and note exactly where it fails without AWS.
  2. Find the smallest change that lets the backend, Celery and Redis run locally with fake AWS keys (the tests already pass with dummy AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY).
  3. Write the working steps as a short section in README.md and open a PR.

Done when

Someone else on the team can follow your steps and get the backend up. Bring a 2-minute summary to the meeting.

Metadata

Metadata

Assignees

Labels

area:docsDocs, onboarding, READMEsarea:trainingDjango + Celery training serviceonboardingFirst task for a new team member

Type

No type

Projects

No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions