Split evaluator

Using Split involves using one of our SDKs. The Split team builds and maintains these SDKs for some of the most popular language libraries and the SDKs are available under open source licenses. For languages where there is no native SDK support, Split offers the Split Evaluator, a small service capable of evaluating some or all available features for a given customer via a REST endpoint.

Setup

The service is available via Docker or command line and its source code is available at https://github.com/splitio/split-evaluator.

Docker (recommended)

• Pull the image: docker pull splitsoftware/split-evaluator

• Run as:

docker run \
  -e SPLIT_EVALUATOR_API_KEY={YOUR_SDK_TYPE_API_KEY} \
  -p 7548:7548 splitsoftware/split-evaluator

Command line

To install the service via command line:

1. Clone the repository: git clone https://github.com/splitio/split-evaluator
2. Prepare the sources: npm install

SPLIT_EVALUATOR_API_KEY=<YOUR_SDK_TYPE_API_KEY> \
SPLIT_EVALUATOR_AUTH_TOKEN=<YOUR_CUSTOM_AUTHENTICATION> \
npm start

Endpoints

The following section will describe the APIs that evaluator can manage. There are grouped in four different resources depending on what they want to achieve.

• Client: it corresponds to the APIs used to send impressions (get treatments) and track events.
• Manager: it corresponds to the APIs that will give you information of the available splits.
• Admin: it corresponds to the APIs that will give you information of the Split Evaluator itself.
• api-docs: it will contain the Swagger specification of Split Evaluator.

/api-docs

Swagger

Split Evaluator uses Swagger to document all API endpoints available in this service. They are available by default in this url: http://localhost:7548/api-docs.
If the port has been modified from the default(7548), make sure to set the environment variable SPLIT_EVALUATOR_SWAGGER_URL to match it.

Client APIs

Corresponds to the Client APIs that is generating impressions and tracking events.

• get-treatment
• get-treatments
• get-treatment-with-config
• get-treatments-with-config
• get-all-treatments
• get-all-treatments-with-config
• track

/client/get-treatment

Evaluates a single feature (split-name) for a single key.

Query params

• key: The key used in the getTreatment call.
• split-name: The name of the split you want to include in the getTreatment call.
• bucketing-key: (Optional) The bucketing key used in the getTreatment call.
• attributes: (Optional) A JSON string of the attributes to include in the getTreatment call of the SDK.

Example

curl 'http://localhost:7548/client/get-treatment?key=my-customer-key&split-name=my-experiment&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "splitName": "my-experiment",
    "treatment": "on"
  }

/client/get-treatments

Provides a way of doing multiple evaluations at once.

Query params

• key: The key used in the getTreatments call.
• split-names: The names of the splits you want to include in the getTreatments call separeted by commas.
• bucketing-key: (Optional) The bucketing key used in the getTreatments call.
• attributes: (Optional) A JSON string of the attributes to include in the getTreatments call of the SDK.

Example

curl 'http://localhost:7548/client/get-treatments?key=my-customer-key&split-names=my-experiment,my-second-experiment&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "my-experiment": {
      "treatment": "on"
    },
    "my-second-experiment": {
      "treatment": "off"
    }
  }

/client/get-treatment-with-config

Evaluates a single feature (split-name) for a single key and adds config in the result.

Query params

• key: The key used in the getTreatmentWithConfig call.
• split-name: The name of the split you want to include in the getTreatmentWithConfig call.
• bucketing-key: (Optional) The bucketing key used in the getTreatmentWithConfig call.
• attributes: (Optional) A JSON string of the attributes to include in the getTreatmentWithConfig call of the SDK.

Example

curl 'http://localhost:7548/client/get-treatment-with-config?key=my-customer-key&split-name=my-experiment&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "splitName": "my-experiment",
    "treatment": "on",
    "config": "{/"color/": /"black/" }"
  }

/client/get-treatments-with-config

Provides a way of doing multiple evaluations at once and attachs configs for each split evaluated.

Query params

• key: The key used in the getTreatmentsWithConfig call.
• split-names: The names of the splits you want to include in the getTreatmentsWithConfig call separeted by commas.
• bucketing-key: (Optional) The bucketing key used in the getTreatmentsWithConfig call.
• attributes: (Optional) A JSON string of the attributes to include in the getTreatmentsWithConfig call of the SDK.

Example

curl 'http://localhost:7548/client/get-treatments-with-config?key=my-customer-key&split-names=my-second-experiment,my-experiment2&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "my-experiment": {
      "treatment": "on",
      "config": "{/"color/": /"black/" }"
    },
    "my-second-experiment": {
      "treatment": "off",
      "config": null
    }
  }

/client/get-all-treatments

Performs multiple evaluations at once. In this case it will match all the splits for a given traffic-type and will perform a getTreatments call with the key provided. You can send more than one {matchingKey,bucketingKey,trafficType} object.

Query params

• keys: The array of keys to be used in the getTreatments call. Each key should specify matchingKey and trafficType. You can also specify bucketingKey.
• attributes: (optional) A JSON string of the attributes to include in the getTreatments call of the SDK.

Example

curl 'http://localhost:7548/client/get-all-treatments?keys=\[\{"matchingKey":"my-first-key","trafficType":"account"\},\{"matchingKey":"my-second-key","bucketingKey":"my-bucketing-key","trafficType":"user"\}\]&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

{
  "account": {
    "my-experiment": {
      "treatment": "on"
    },
    "my-second-experiment": {
      "treatment": "off"
    }
  },
  "user": {
    "my-last-experiment": {
      "treatment": "off"
    }
  }
}

/client/get-all-treatments-with-config

Performs multiple evaluations at once. In this case it will match all the splits for a given traffic-type and will perform a getTreatmentsWithConfig call with the key provided. You can send more than one {matchingKey,bucketingKey,trafficType} object. This endpoint will also adds the configurations for particular split.

Query params

• keys: The array of keys to be used in the getTreatmentsWithConfig call. Each key should specify matchingKey and trafficType. You can also specify bucketingKey.
• attributes: (optional) A JSON string of the attributes to include in the getTreatmentsWithConfig call of the SDK.

Example

curl 'http://localhost:7548/client/get-all-treatments-with-config?keys=\[\{"matchingKey":"my-first-key","trafficType":"account"\},\{"matchingKey":"my-second-key","bucketingKey":"my-bucketing-key","trafficType":"user"\}\]&attributes=\{"attribute1":"one","attribute2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

{
  "account": {
    "my-experiment": {
      "treatment": "on",
       "config": "{/"color/": /"black/" }"
    },
    "my-second-experiment": {
      "treatment": "off",
      "config": null
    }
  },
  "user": {
    "my-last-experiment": {
      "treatment": "off",
      "config": "{/"copy/": /"better copy/" }"
    }
  }
}

/client/track

Records any actions your customers perform. Each action is known as an event and corresponds to an event type. Calling track allows you to measure the impact of your splits on your users' actions and metrics.

Query params

• key: The key used in the track call.
• traffic-type: The traffic type of the key that you want to include in the track call.
• event-type: The event type that this event should correspond to the track call of the SDK.
• value: (Optional) value to be used in creating the metric.
• properties: (Optional) A JSON string of the properties to include in the track call of the SDK for filtering your metrics.

Example

curl 'http://localhost:7548/client/track?key=my-customer-key&event-type=my-event&traffic-type=account&properties=\{"prop1":"one","prop2":2,"attribute3":true\}' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

Successfully queued event

Manager APIs

Provides information of all the available splits.

• split
• splits
• names

/manager/split

Provides information of one particular split.

Query params

• split-name: The name of the split you want to have information. It uses split call of the SDK.

Example

curl 'http://localhost:7548/manager/split?split-name=my-experiment' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "name": "my-experiment",
    "trafficType": "user",
    "killed": false,
    "changeNumber": 1563394983932,
    "treatments": ["on", "off"],
    "configs": {
      "on": "{/"color/": /"black/" }",
      "off": "{/"color/": /"red/" }"
    }
  }

/manager/splits

Provides information of all the available splits by calling splits method from the SDK.

Example

curl 'http://localhost:7548/manager/splits' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "splits": [{
      "name": "my-experiment",
      "trafficType": "user",
      "killed": false,
      "changeNumber": 1563394983932,
      "treatments": ["on", "off"],
      "configs": {
        "on": "{/"color/": /"black/" }",
        "off": "{/"color/": /"red/" }"
      }
    }, {
      "name": "my-second-experiment",
      "trafficType": "user",
      "killed": false,
      "changeNumber": 1563394983933,
      "treatments": ["on", "off"],
      "configs": {
        "on": "{/"color/": /"yellow/" }",
        "off": "{/"color/": /"green/" }"
      }
    }]
  }

/manager/names

Provides the names of the available splits. It calls names of the SDK.

Example

curl 'http://localhost:7548/manager/names' -H 'Authorization: {SPLIT_EVALUATOR_AUTH_TOKEN}'

Response

  {
    "splits": ["my-experiment", "my-second-experiment"]
  }

Admin

Provides information about Split Evaluator itself.

• ping
• uptime
• healthcheck
• version
• machine

/admin/ping

A ping endpoint to monitor the service status. If the service is running, it responds with pong and the HTTP status code 200.

Example

curl 'http://localhost:7548/admin/ping'

/admin/uptime

Returns the uptime of the service in a human-readable string.

Example

curl 'http://localhost:7548/admin/uptime'

Response

"5d 3h 36m 39s"

/admin/healthcheck

Checks that everything is working as expected before sending evaluations. It returns a status code of either 200 or 500, depending on the result of the check, along with a message explaining the status.

Example

curl 'http://localhost:7548/admin/healthcheck'

Response

// For 200 status code
"Split Evaluator working as expected."
// For 500 status code
"Split evaluator engine not evaluating traffic properly."

/admin/version

Version information for the evaluator and the SDK within.

Example

curl 'http://localhost:7548/admin/version'

Response

  {
    "version": "1.0.2",
    "sdk": "nodejs",
    "sdkVersion": "9.3.4"
  }

/admin/machine

Returns the data for the machine where this service is running.

Example

curl 'http://localhost:7548/admin/machine'

Response

  {
    "ip": "10.0.0.125",
    "name": "machine_name"
  }

Impression Listener

The Split Evaluator provides an impression listener (SPLIT_EVALUATOR_IMPRESSION_LISTENER_ENDPOINT) that bulks post impressions to a user-defined HTTP endpoint. The endpoint should expect a POST request, containing a JSON body with the following format.

If an Impression Listener is provided when the Split Evaluator is initialized, a task will run in background that will post impressions. There will be two ways of posting impressions to the provided endpoint:
1) Every 30 seconds by the Evaluator.
2) When the queue of impressions reached the max amount of impressions.

For further information, about how configuring the Impression Listener, see Configuration.

  {
    "impressions": [{
      "feature": "my-experiment",
      "keyName": "key-experiment",
      "treatment": "on",
      "time": 987654321,
      "changeNumber": 987654321,
      "label": "label"
    }, {
      "feature": "my-experiment",
      "keyName": "key-experiment",
      "treatment": "off",
      "time": 123456789,
      "changeNumber": 123456789,
      "label": "default rule"
    }
  }

Configuration

The available configuration variables are listed below. Always use -e <VARIABLE>=<VALUE> with Docker.

HTTPS/SSL

This service does not currently support a secured connection. We recommend running this service in a redundant manner behind a load balancer such as AWS ELB or Nginx, with SSL termination.

Reach out to us any time at support@split.io if you have questions.