Skip to main content

Ruby

Eppo's Ruby SDK can be used for both feature flagging and experiment assignment:

1. Install the SDK

Install the SDK with gem:

gem install eppo-server-sdk

2. Initialize the SDK

Initialize the SDK with an API key, which can be generated in the Eppo interface. Initialization the SDK when your application starts up to generate a singleton client instance, once per application lifecycle:

require 'eppo_client'
from eppo_client.config import Config

config = EppoClient::Config.new('<YOUR_API_KEY>')
client = EppoClient::init(config)

After initialization, the SDK begins polling Eppo’s API at regular intervals to retrieve the most recent experiment configurations such as variation values and traffic allocation. The SDK stores these configurations in memory so that assignments are effectively instant. If you are using the SDK for experiment assignments, make sure to pass in an assignment logging callback (see section below).

Define an assignment logger (experiment assignment only)

If you are using the Eppo SDK for experiment assignment (i.e randomization), include a logger instance in the config that is passed to the init function on SDK initialization. The SDK invokes the log_assignment method in the instance to capture assignment data whenever a variation is assigned.

The code below illustrates an example implementation of logging with Segment, but you could also use other event-tracking systems. The only requirement is that the SDK can call a log_assignment method. Here we override Eppo's AssignmentLogger class with a function named log_assignment, then instantiate a config using an instance of the custom logger class, and finally instantiate the client:

require 'segment/analytics'

# Connect to Segment (or your own event-tracking system)
Analytics = Segment::Analytics.new({ write_key: 'SEGMENT_WRITE_KEY' })

class CustomAssignmentLogger < EppoClient::AssignmentLogger
def log_assignment(assignment)
Analytics.track(assignment["subject"], "Eppo Assignment", assignment)
end
end

config = EppoClient::Config.new(
'<YOUR_API_KEY>',
assignment_logger: CustomAssignmentLogger.new
)
client = EppoClient::init(config)

The SDK will invoke the log_assignment function with an assignment object that contains the following fields:

FieldDescriptionExample
experiment (string)An Eppo experiment key"recommendation_algo"
subject (string)An identifier of the subject or user assigned to the experiment variationUUID
variation (string)The experiment variation the subject was assigned to"control"
timestamp (string)The time when the subject was assigned to the variation2021-06-22T17:35:12.000Z
subjectAttributes (map)A free-form map of metadata about the subject. These attributes are only logged if passed to the SDK assignment function{ "country": "US" }
note

More examples of logging (with Segment, Rudderstack, mParticle, and Snowplow) can be found in the event logging page.

3. Assign variations

Assigning users to flags or experiments with a single get_assignment function:

require 'eppo_client'

client = EppoClient::Client.instance
variation = client.get_assignment(
'<SUBJECT-KEY>',
'<FLAG-OR-EXPERIMENT-KEY>',
{
# Optional map of subject metadata for targeting.
}
)

The get_assignment function takes two required and one optional input to assign a variation:

  • subject_key - The entity ID that is being experimented on, typically represented by a uuid.
  • flag_or_experiment_key - This key is available on the detail page for both flags and experiments.
  • subject_attributes - An optional map of metadata about the subject used for targeting. If you create rules based on attributes on a flag/experiment, those attributes should be passed in on every assignment call.

Handling nil

We recommend always handling the nil case in your code. Here are some examples of when the SDK returns nil:

  1. The Traffic Exposure setting on experiments/allocations determines the percentage of subjects the SDK will assign to that experiment/allocation. For example, if Traffic Exposure is 25%, the SDK will assign a variation for 25% of subjects and nil for the remaining 75% (unless the subject is part of an allow list).

  2. If you are using Eppo for experiment assignments, you must start the experiment in the user interface before get_assignment returns variations. It will return nil if the experiment is not running, both before and after.

    start-experiment

  3. If get_assignment is invoked before the SDK has finished initializing, the SDK may not have access to the most recent experiment configurations. In this case, the SDK will assign a variation based on any previously downloaded experiment configurations stored in local storage, or return nil if no configurations have been downloaded.

Debugging nil

If you need more visibility into why get_assignment is returning nil, you can change the logging level to Logger::DEBUG to see more details in the standard output.

require 'eppo_client'
require 'logger'

client = EppoClient::Client.instance
variation = client.get_assignment(
'<SUBJECT-KEY>',
'<FLAG-OR-EXPERIMENT-KEY>',
{},
Logger::DEBUG
)

note

It may take up to 10 seconds for changes to Eppo experiments to be reflected by the SDK assignments.