This is the Alpaca official docker container packages zipline. With this container image, you can easily run your Quantopian algorithm with live trading. This is to run your algorithm in your computer, or server by yourself, and Alpaca is also desiging a "one-click" solution to run your algorithm without having your server.
All you need is docker installed in your system.
There are two usages for this container. One is a Jupyter notebook enivronment for the exploration purpose, and the other is a CLI environment to launch your algorithm for real money trading.
In either case, you should set Alpaca API Key environment variables first.
$ export APCA_API_KEY_ID=xxxxx
$ export APCA_API_SECRET_KEY=yyyy
The container installs Alpaca's official python SDK.
By default, the container launches a Jupyter notebook with no password on the port 8888. You need to forward the port so you can access the notebook from your browser.
$ docker run -it --rm \
-e APCA_API_KEY_ID -e APCA_API_SECRET_KEY -p 8888:8888 \
alpacamarkets/alpaca-zipline
Then, open http://localhost:8888/ . You can start accessing Alpaca API immediately on this environment.
You can start a CLI environment using the same image as below.
# some data directory to download the bundle data
$ DATADIR=/data
# the directory where your algorithm file is located
$ ALGODIR=/algorithm
$ docker run -it --rm -e APCA_API_KEY_ID -e APCA_API_SECRET_KEY \
-v $DATADIR:/root/.zipline/data \
-v $ALGODIR:/work -w /work \
alpacamarkets/alpaca-zipline bash
Modify DATADIR and ALGODIR for your needs. The data directory is used to store the
bundle data and better to be mounted from the host directory. The algorithm directory
is where our algorithm file is located.
In order to run the live trading, you still need to have a bundle data. Since zipline's default quantopian-quandl bundle covers very limited data (3k stocks with daily only), Alpaca offers more realistic data bundle called alpaca-bundle (8k stocks with daily/minute last 3000 bars). From inside the container, simply run:
$ zipline ingest -b alpaca-bundle
It may take a couple of minutes to complete.
Once the bundle is saved, you are ready to run your algorithm.
$ zipline run -f /work/myalgo.py \
--broker=alpaca --broker-uri=dummy \
--state-file /work/state \
--realtime-bar-target /work/realtime-bars/ \
--bundle alpaca-bundle --data-frequency minute
Note the --broker-uri parameter is ignored with alpaca broker, but you have to specify something.
The --state-file is the file location where the zipline context is serialized.
Quantopian runs each algorithm in Python 2, but this container is packaged in Python 3. You may need to adjust some of your code for Python 3. You will also need to insert this code at the beginning of your algorithm file that fills Quantopian's implicit import modules.
from zipline.api import *
import logbook
log = logbook.Logger('algo')
While we have verified a few algorithms runnable with this image, we are aware of many different use cases, and we are happy to hear your issues, and please let us know. Alpaca constantly keeps improving the image so that it covers more use cases.