Sixpack 1.0.2

Requirements

• Redis >= 2.6

• Python >= 2.7 (3.0 untested, pull requests welcome)

Getting Started

To get going, create (or don’t, but you really should) a new virtualenv for your sixpack installation. Follow that up with pip install:

$ pip install sixpack

Note: If you get an error like src/hiredis.h:4:20: fatal error: Python.h: No such file or directory you need to install the python development tools. apt-get install python-dev on Ubuntu.

Next, create a Sixpack configuration. A configuration must be created for sixpack to run. Here’s the default:

redis_port: 6379                        # Redis port
redis_host: localhost                   # Redis host
redis_prefix: sixpack                   # all Redis keys will be prefixed with this
redis_db: 15                            # DB number in redis

enable_whiplash: False                  # Enable the whiplash/multi-armed bandit alternative algorithm

# The regex to match for robots
robot_regex: $^|trivial|facebook|MetaURI|butterfly|google|amazon|goldfire|sleuth|xenu|msnbot|SiteUptime|Slurp|WordPress|ZIBB|ZyBorg|pingdom|bot|yahoo|slurp|java|fetch|spider|url|crawl|oneriot|abby|commentreader|twiceler
ignored_ip_addresses: []                # List of IP

asset_path: gen                         # Path for compressed assets to live. This path is RELATIVE to sixpack/static
secret_key: '<your secret key here>'    # Random key (any string is valid, required for sixpack-web to run)

You can store this file anywhere (we recommend /etc/sixpack/config.yml). As long as Redis is running, you can now start the sixpack server like this:

$ SIXPACK_CONFIG=<path to config.yml> sixpack

Sixpack-server will be listening on port 5000 by default but can be changed with the SIXPACK_PORT environment variable. For use in a production environment, please see the “Production Notes” section below.

Using the API

All interaction with Sixpack is done via HTTP GET requests. Sixpack allows for cross-language testing by accepting a unique client_id (which the client is responsible for generating) that links a participation to a conversion. All requests to Sixpack require a client_id.

$ curl http://localhost:5000/participate?experiment=button_color&alternatives=red&alternatives=blue&client_id=12345678-1234-5678-1234-567812345678

$ curl http://localhost:5000/participate?experiment=button_color&alternatives=red&alternatives=blue&force=red&client_id=12345678-1234-5678-1234-567812345678

{
    status: "ok",
    alternative: {
        name: "red"
    },
    experiment: {
        name: "button_color"
    },
    client_id: "12345678-1234-5678-1234-567812345678"
}

$ curl http://localhost:5000/convert?experiment=button_color&client_id=12345678-1234-5678-1234-567812345678

Clients

We’ve already provided clients in four languages. We’d love to add clients in additional languages. If you feel inclined to create one, please first read the CLIENTSPEC. After writing your client, please update and pull request this file so we know about it.

• Ruby

• Python

• JavaScript

• PHP

• Go

Dashboard

Sixpack comes with a built in dashboard. You can start the dashboard with:

$ SIXPACK_CONFIG=<path to config.yml> sixpack-web

The sixpack dashboard allows you to visualize how each experiment’s alternatives are doing compared to the rest, select alternatives as winners, and update experiment descriptions to something more human-readable

Sixpack-web defaults to run on port 5001 but can be changed with the SIXPACK_WEB_PORT environment variable.

$ curl http://localhost:5001/experiments.json

$ curl http://localhost:5001/experiments/blue-or-red-header.json

Production Notes

We recommend running Sixpack on gunicorn in production. You will need to install gunicorn in your virtual environment before running the following.

To run the sixpack server using gunicorn/gevent (a separate installation) you can run the following:

gunicorn --access-logfile - -w 8 --worker-class=gevent sixpack.server:start

To run the sixpack web dashboard using gunicorn/gevent (a separate installation) you can run the following:

gunicorn --access-logfile - -w 2 --worker-class=gevent sixpack.web:start

Note: After selecting an experiment winner, it is best to remove the Sixpack experiment code from your codebase to avoid unnecessary http requests.

Contributing

1. Fork it

2. Start Sixpack in development mode with

$ PYTHONPATH=. SIXPACK_CONFIG=<path to config.yml> bin/sixpack

and

$ PYTHONPATH=. SIXPACK_CONFIG=<path to config.yml> bin/sixpack-web

We’ve also included a small script that will seed Sixpack with lots of random data for testing and development on sixpack-web. You can seed Sixpack with the following command

$ PYTHONPATH=. SIXPACK_CONFIG=<path to config.yml> sixpack/test/seed

This command will make a few dozen requests to the participate and convert endpoints. Feel free to run it multiple times to get additional data.

Note: By default the server runs in production mode. If you’d like to turn on Flask and Werkzeug debug modes set the SIXPACK_DEBUG environment variable to true.

3. Create your feature branch (git checkout -b my-new-feature)

4. Write tests

5. Run tests with nosetests

6. Commit your changes (git commit -am 'Added some feature')

7. Push to the branch (git push origin my-new-feature)

8. Create new pull request

Please avoid changing versions numbers; we’ll take care of that for you

License

Sixpack is released under the BSD 2-Clause License.