greybox Wrapped is a project that allows us to display statistics and achievements from academic debates. As the Czech Debating Association, we track all academic debates that take place at our events. From this database, data can be extracted to analyse performance and award individuals with achievements. This project will allow us to analyse the data and award individuals' achievements on the backend. On the frontend, individuals will be able to request personalised presentation of their achievements each year, just as other applications do a "rewind" of the year. Sometimes in the code, greybox wrapped can be called ADK wrapped, which is the same thing, only an older name.
We're using a standard client-server architecture with a stationary pre-filled backend database.
sql_prep/
: A training component that restructures the data into a series of achievements / presentable statistics. This is handled using dbt/SQL with a sprinkle of Python on top of a DuckDB database, since our dataset is small (<50MB).backend/
: A transfer component that communicates the data to the presentation layer via API. FastAPI handles this.frontend/
: A presentation component that displays the data in a meaningful/beautiful way. React/Spectacle handles this.
graph TD
subgraph Meltano Data Load
DB1(MySQL Remote) -->|Exports data to| Dump(SQL Dump)
Dump -->|Loads data into| DB2(MySQL Local)
DB2 -->|Loads data via Meltano into| C(DuckDB)
end
subgraph Frontend Container
A(Spectacle/React)
end
subgraph Backend Container
A(Spectacle/React) -->|Requests data from| B(FastAPI)
B(FastAPI) -->|Sends data to| A(Spectacle/React)
B(FastAPI) -->|Processes request and retrieves data from| C(DuckDB)
C(DuckDB) -->| Sends data to| B(FastAPI)
end
subgraph dbt
D(dbt/SQL) -->| Ahead of time, prepares views and tables in | C(DuckDB)
end
- Get a DuckDB export of the dataset and place it in
data/adk_wrapped.db
. - Create a
profiles.yml
file in~/.dbt
with the following contents:adk_wrapped: outputs: dev: path: "path/to/data/adk_wrapped.db" schema: adk_wrapped type: duckdb threads: 4 extensions: - httpfs - parquet target: dev
- Optionally, get a copy of the rainbow tables and place them in
data/rainbow_tables.csv
. Ensure that the column names aregreybox_id
andhash
. - For local development, run
make setup
to install the dependencies, including and especiallydbt-duckdb
. (This doesn't set up the frontend, though.)- Run
make run
to letdbt
populate the database with transformations. - Run
make backend
to start the FastAPI server. - In a separate terminal, run
make frontend
to start the React server. (To be able to do that, you'll need to set up Node Version Manager for Windows -nvm
first and, using it, Node v18.15.0.)
- Run
- For deployment/testing,
docker-compose up
should do everything.
All endpoints are managed by nginx, which is packaged in frontend/Dockerfile
.
/
displays a prompt to go back to Greybox 2.0 and follow the requisite link./slides/{person_id}
to display the achievements of a given person./link/{id_hash}
to display the achievements of a given person, with a lookup provided via a rainbow table entry.
/api
to display the API documentation./api/achievements/{greybox_id}
to pass the achievements to the presentation layer./api/link/{id_hash}
to pass the achievements to the presentation layer, with a lookup provided via a rainbow table entry.
We're using fly.io to deploy the backend and frontend, both using the Dockerfile
and fly.toml
in the respective directories. We're basically playing the role of docker-compose.yml
by manually handling the volumes, networking, and environment variables.
The fly.toml
files are already set up now. The prerequisites were:
cd frontend
fly apps create adk-wrapped
cd ../backend
fly apps create adk-wrapped-api
fly volumes create adk_wrapped_data --region ams --size
The app is currently deployed to the Amsterdam (ams
) region.
make deploy
pushes both the frontend and backend to fly.io. However, to update the volume, we must ssh/SFTP into the backend nad puts
it there manually.
cd backend
fly ssh console
# SSH prompt will appear
rm /data/adk_wrapped.db
exit
fly ssh sftp shell
# SFTP prompt will appear
cd /data
puts ../data/adk_wrapped.db
# press Ctrl+D to exit
Before you can do either, you need to be a member of the deploying organization, which is Simon's personal one, and install fly CLI + authenticate.