This is an experimental setup with simple goal to tryout Apache Superset as Trading dashboard + simple implementation of Reinforcement learning approach for algo-trading with Bitcoin. If you have an idea for how to add more hype to the pile, open an issue!
Setup consists of 3 main modules:
-
trader
: Trading scripts for fetching historical Bitcoin prices and backtesting and live (paper) trading on Bitstamp, with simple Q-table RL algo-trader implemented withtensorflow
andpyalgotrade
:$$\mathbf{b}$$ -
airflow
: Apache Airflow DAGs for scheduling initial and weekly tasks. -
superset
: Apache Superset Dashboard
First, make superset-config.env
and trader-config.env
files based on examples and populate keys
(no spaces around = sign and no quotes).
Build docker images:
docker-compose build
docker-compose up
or
docker-compose up --build
In order to make or load Dashboard you'll need to prepare database, load it to Apache Superset and import Dashboard pickle file.
- Go to Airflow UI (
localhost:8090
) and turn-onweekly
DAG. - After DAG finishes, go to Superset UI (
localhost:8088
) and login with credentials from env file. - Import database: internally it is mounted to
/etc/superset/db
and calledtrader.db
. - Import pickled Dashboard from superset/dashboard directory.
- Make some changes (add other traders, tweak RLtrader, etc..)
- Deploy somewhere, make guest user and invite people to show off your Dashboards!