Keen IO Official Ruby Client Library

Build Status Code Climate Gem Version

keen-gem is the official Ruby Client for the Keen IO API. The Keen IO API lets developers build analytics features directly into their apps.


Add to your Gemfile:

gem 'keen'

or install from Rubygems:

gem install keen

keen is tested with Ruby 1.9.3 + and on:

  • MRI
  • Rubinius
  • jRuby (except for asynchronous methods - no TLS support for EM on jRuby)


Before making any API calls, you must supply keen-gem with a Project ID and one or more authentication keys. (If you need a Keen IO account, sign up here - it's free.)

Setting a write key is required for publishing events. Setting a read key is required for running queries. Setting a master key is required for performing deletes. You can find keys for all of your projects on

The recommended way to set keys is via the environment. The keys you can set are KEEN_PROJECT_ID, KEEN_WRITE_KEY, KEEN_READ_KEY and KEEN_MASTER_KEY. You only need to specify the keys that correspond to the API calls you'll be performing. If you're using foreman, add this to your .env file:


If not, make a script to export the variables into your shell or put it before the command you use to start your server.

When you deploy, make sure your production environment variables are set. For example, set config vars on Heroku. (We recommend this environment-based approach because it keeps sensitive information out of the codebase. If you can't do this, see the alternatives below.)

Once your environment is properly configured, the Keen object is ready to go immediately.

Synchronous Publishing

Publishing events requires that KEEN_WRITE_KEY is set. Publish an event like this:

Keen.publish(:sign_ups, { :username => "lloyd", :referred_by => "harry" })

This will publish an event to the sign_ups collection with the username and referred_by properties set. The event properties can be any valid Ruby hash. Nested properties are allowed. Lists of objects are also allowed, but not recommended because they can be difficult to query over. See alternatives to lists of objects here. You can learn more about data modeling with Keen IO with the Data Modeling Guide.

Protip: Marshalling gems like Blockhead make converting structs or objects to hashes easier.

The event collection need not exist in advance. If it doesn't exist, Keen IO will create it on the first request.

Asynchronous publishing

Publishing events shouldn't slow your application down or make users wait longer for page loads & server requests.

The Keen IO API is fast, but any synchronous network call you make will negatively impact response times. For this reason, we recommend you use the publish_async method to send events when latency is a concern. Alternatively, you can drop events into a background queue e.g. Delayed Jobs and publish synchronously from there.

To publish asynchronously, first add em-http-request to your Gemfile. Make sure it's version 1.0 or above.

gem "em-http-request", "~> 1.0"

Next, run an instance of EventMachine. If you're using an EventMachine-based web server like thin or goliath you're already doing this. Otherwise, you'll need to start an EventMachine loop manually as follows:

require 'em-http-request' { }

The best place for this is in an initializer, or anywhere that runs when your app boots up. Here's a useful blog article that explains more about this approach - EventMachine and Passenger.

And here's a gist that shows an example of Eventmachine with Unicorn, specifically the Unicorn config for starting and stopping EventMachine after forking.

Now, in your code, replace publish with publish_async. Bind callbacks if you require them.

http = Keen.publish_async("sign_ups", { :username => "lloyd", :referred_by => "harry" })
http.callback { |response| puts "Success: #{response}"}
http.errback { puts "was a failurrr :,(" }

This will schedule the network call into the event loop and allow your request thread to resume processing immediately.

Running queries

The Keen IO API provides rich querying capabilities against your event data set. For more information, see the Data Analysis API Guide.

Running queries requires that KEEN_READ_KEY is set.

Here are some examples of querying with keen-gem. Let's assume you've added some events to the "purchases" collection.

# Various analysis types
Keen.count("purchases") # => 100
Keen.sum("purchases", :target_property => "price", :timeframe => "today")  # => 10000
Keen.minimum("purchases", :target_property => "price", :timeframe => "today")  # => 20
Keen.maximum("purchases", :target_property => "price", :timeframe => "today")  # => 100
Keen.average("purchases", :target_property => "price", :timeframe => "today")  # => 60
Keen.median("purchases", :target_property => "price", :timeframe => "today")  # => 60
Keen.percentile("purchases", :target_property => "price", :percentile => 90, :timeframe => "today")  # => 100
Keen.count_unique("purchases", :target_property => "username", :timeframe => "today")  # => 3
Keen.select_unique("purchases", :target_property => "username", :timeframe => "today")  # => ["Bob", "Linda", "Travis"]

# Group by's and filters
Keen.sum("purchases", :target_property => "price", :group_by => "", :timeframe => "this_14_days")  # => [{ "": 123, "result": 240 }]
Keen.count("purchases", :timeframe => "today", :filters => [{
    "property_name" => "referred_by",
    "operator" => "eq",
    "property_value" => "harry"
  }]) # => 2

# Relative timeframes
Keen.count("purchases", :timeframe => "today") # => 10

# Absolute timeframes
Keen.count("purchases", :timeframe => {
  :start => "2015-01-01T00:00:00Z",
  :end => "2015-31-01T00:00:00Z"
}) # => 5

# Extractions
Keen.extraction("purchases", :timeframe => "today")  # => [{ "keen" => { "timestamp" => "2014-01-01T00:00:00Z" }, "price" => 20 }]

# Funnels
Keen.funnel(:steps => [{
  :actor_property => "username", :event_collection => "purchases", :timeframe => "yesterday" }, {
  :actor_property => "username", :event_collection => "referrals", :timeframe => "yesterday" }]) # => [20, 15]

# Multi-analysis
Keen.multi_analysis("purchases", analyses: {
  :gross =>      { :analysis_type => "sum", :target_property => "price" },
  :customers =>  { :analysis_type => "count_unique", :target_property => "username" } },
  :timeframe => 'today', :group_by => "") # => [{ "" => 2, "gross" => 314.49, "customers" => 8 } }]

Many of these queries can be performed with group by, filters, series and intervals. The response is returned as a Ruby Hash or Array.

Detailed information on available parameters for each API resource can be found on the API Technical Reference.

The Query Method

You can also specify the analysis type as a parameter to a method called query:

Keen.query("median", "purchases", :target_property => "price")  # => 60

This simplifes querying code where the analysis type is dynamic.

Query Options

Each query method or alias takes an optional hash of options as an additional parameter. Possible keys are:

:response – Set to :all_keys to return the full API response (usually only the value of the "result" key is returned). :method - Set to :post to enable post body based query (

Query Logging

You can log all GET and POST queries automatically by setting the log_queries option.

Keen.log_queries = true
# I, [2016-10-30T11:45:24.678745 #9978]  INFO -- : [KEEN] Send GET query to<YOUR_PROJECT_ID>/queries/count?event_collection=purchases with options {}

Saved Queries

You can manage your saved queries from the Keen ruby client.

# Create a saved query
Keen.saved_queries.create("name", saved_query_attributes)

# Get all saved queries

# Get one saved query

# Get saved query with results
Keen.saved_queries.get("saved-query-slug", results: true)

# Update a saved query
saved_query_attributes = { refresh_rate: 14400 }
Keen.saved_queries.update("saved-query-slug", saved_query_attributes)

# Delete a saved query
Getting Query URLs

Sometimes you just want the URL for a query, but don't actually need to run it. Maybe to paste into a dashboard, or open in your browser. In that case, use the query_url method:

Keen.query_url("median", "purchases", :target_property => "price", { :timeframe => "today" })
# => "<project-id>/queries/median?target_property=price&event_collection=purchases&api_key=<api-key>"

If you don't want the API key included, pass the :exclude_api_key option:

Keen.query_url("median", "purchases", { :target_property => "price", :timeframe => "today" }, :exclude_api_key => true)
# => "<project-id>/queries/median?target_property=price&event_collection=purchases"

Listing collections

The Keen IO API let you get the event collections for the project set, it includes properties and their type. It also returns links to the collection resource.

Keen.event_collections # => [{ "name": "purchases", "properties": { "": "num", ... }, ... }]

Getting the list of event collections requires that the KEEN_MASTER_KEY is set.

Deleting events

The Keen IO API allows you to delete events from event collections, optionally supplying a filter to narrow the scope of what you would like to delete.

Deleting events requires that the KEEN_MASTER_KEY is set.

# Assume some events in the 'signups' collection

# We can delete them all
Keen.delete(:signups)  # => true

# Or just delete an event corresponding to a particular user
Keen.delete(:signups, filters: [{
  :property_name => 'username', :operator => 'eq', :property_value => "Bob"
}])  # => true

Other code examples

Overwriting event timestamps

Two time-related properties are included in your event automatically. The properties “keen.timestamp” and “keen.created_at” are set at the time your event is recorded. You have the ability to overwrite the keen.timestamp property. This could be useful, for example, if you are backfilling historical data. Be sure to use ISO-8601 Format.

Keen stores all date and time information in UTC!

Keen.publish(:sign_ups, {
  :keen => { :timestamp => "2012-12-14T20:24:01.123000+00:00" },
  :username => "lloyd",
  :referred_by => "harry"

Batch publishing

The keen-gem supports publishing events in batches via the publish_batch method. Here's an example usage:

  :signups => [
    { :name => "Bob" },
    { :name => "Mary" }
  :purchases => [
    { :price => 10 },
    { :price => 20 }

This call would publish 2 signups events and 2 purchases events - all in just one API call. Batch publishing is ideal for loading historical events into Keen IO.

Asynchronous batch publishing

Ensuring the above guidance is followed for asynchronous publishing, batch publishing logic can used asynchronously with publish_batch_async:

  :signups => [
    { :name => "Bob" },
    { :name => "Mary" }
  :purchases => [
    { :price => 10 },
    { :price => 20 }

Configurable and per-client authentication

To configure keen-gem in code, do as follows:

Keen.project_id = 'xxxxxxxxxxxxxxx'
Keen.write_key = 'yyyyyyyyyyyyyyy'
Keen.read_key = 'zzzzzzzzzzzzzzz'
Keen.master_key = 'aaaaaaaaaaaaaaa'

You can also configure unique client instances as follows:

keen = => 'xxxxxxxxxxxxxxx',
                        :write_key  => 'yyyyyyyyyyyyyyy',
                        :read_key   => 'zzzzzzzzzzzzzzz',
                        :master_key => 'aaaaaaaaaaaaaaa')


keen-gem can be used with em-synchrony. If you call publish_async and EM::Synchrony is defined the method will return the response directly. (It does not return the deferrable on which to register callbacks.) Likewise, it will raise exceptions 'synchronously' should they happen.

Beacon URLs

It's possible to publish events to your Keen IO project using the HTTP GET method. This is useful for situations like tracking email opens using image beacons.

In this situation, the JSON event data is passed by encoding it base-64 and adding it as a request parameter called data. The beacon_url method found on the Keen::Client does this for you. Here's an example:

Keen.project_id = 'xxxxxx';
Keen.write_key = 'yyyyyy';
Keen.beacon_url("sign_ups", :recipient => "")
  # => ""

To track email opens, simply add an image to your email template that points to this URL. For further information on how to do this, see the image beacon documentation.

Redirect URLs

Redirect URLs are just like image beacon URLs with the addition of a redirect query parameter. This parameter is used to issue a redirect to a certain URL after an event is recorded.

Keen.redirect_url("sign_ups", { :recipient => "" }, "")
  # => ""

This is helpful for tracking email clickthroughs. See the redirect documentation for further information.

Generating scoped keys

Note, Scoped Keys are now deprecated in favor of access keys.

A scoped key is a string, generated with your API Key, that represents some encrypted authentication and query options. Use them to control what data queries have access to.

# "my-api-key" should be your MASTER API key
scoped_key ="my-api-key", { "filters" => [{
  "property_name" => "accountId",
  "operator" => "eq",
  "property_value" => "123456"
}]}).encrypt! # "4d1982fe601b359a5cab7ac7845d3bf27026936cdbf8ce0ab4ebcb6930d6cf7f139e..."

You can use the scoped key created in Ruby for API requests from any client. Scoped keys are commonly used in JavaScript, where credentials are visible and need to be protected.

Access Keys

You can use access keys to restrict the functionality of a key you use with the Keen API. Access keys can also enrich events that you send.

Read up on the full key body options.

Create a key that automatically adds information to each event published with that key:

key_body = {
  "name" => "autofill foo",
  "is_active" => true,
  "permitted" => ["writes"],
  "options" => {
    "writes" => {
      "autofill": {
        "foo": "bar"

new_key = client.access_keys.create(key_body)
autofill_write_key = new_key["key"]

You can revoke and unrevoke keys to disable or enable access. all will return all current keys for the project, while get("key-value-here") will return info for a single key. You can also update and delete keys.

Additional options

HTTP Read Timeout

The default Net::HTTP timeout is 60 seconds. That's usually enough, but if you're querying over a large collection you may need to increase it. The timeout on the API side is 300 seconds, so that's as far as you'd want to go. You can configure a read timeout (in seconds) by setting a KEEN_READ_TIMEOUT environment variable, or by passing in a read_timeout option to the client constructor as follows:

keen = => 300)

You can also configure the NET::HTTP open timeout, default is 60 seconds. To configure the timeout (in seconds) either set KEEN_OPEN_TIMEOUT environment variable, or by passing in a open_timeout option to the client constructor as follows:

keen = => 30)
HTTP Proxy

You can set the KEEN_PROXY_TYPE and KEEN_PROXY_URL environment variables to enable HTTP proxying. KEEN_PROXY_TYPE should be set to socks5. You can also configure this on client instances by passing in proxy_type and proxy_url keys.

keen = => 'socks5', :proxy_url => 'http://localhost:8888')



If you run into Keen::Error: Keen IO Exception: An EventMachine loop must be running to use publish_async calls or Uncaught RuntimeError: eventmachine not initialized: evma_set_pending_connect_timeout, this means that the EventMachine loop has died. This can happen for a variety of reasons, and every app is different. Issue #22 shows how to add some extra protection to avoid this situation.

publish_async in a script or worker

If you write a script that uses publish_async, you need to keep the script alive long enough for the call(s) to complete. EventMachine itself won't do this because it runs in a different thread. Here's an example gist that shows how to exit the process after the event has been recorded.

Additional Considerations


It's not just us humans that browse the web. Spiders, crawlers, and bots share the pipes too. When it comes to analytics, this can cause a mild headache. Events generated by bots can inflate your metrics and eat up your event quota.

If you want some bot protection, check out the Voight-Kampff gem. Use the gem's method to detect bots and avoid logging events.


  • Added an option to log queries
  • Added a cli option that includes the Keen code
  • Add support for access keys
  • Move saved queries into the Keen namespace
  • Deprecate scoped keys in favor of access keys
  • Remove support for ruby 1.9.3
  • Update a few dependencies
  • Add ability to set the open_time setting for the http client.
  • Added the ability to send additional optional headers.
  • Added a new header Keen-Sdk that sends the SDK version information on all requests.
  • Fix bug with scoped key generation not working with newer Keen projects.
  • Add SDK support for Saved Queries
  • Removed support for Ruby MRI 1.8.7
  • Added support for max_age as an integer.
  • Added support for setting an IV for scoped keys. Thanks @anatolydwnld
  • Add support for a configurable read timeout
  • Add support for returning all keys back from query API responses
  • Add support for getting query URLs
  • Make the query method public so code supporting dynamic analysis types is easier to write
  • Add support for median and percentile analysis
  • Support arrays for extraction property_names option
  • Add support for asynchronous batch publishing
  • UPGRADE WARNING Do you use spaces in collection names? Or other special characters? Read this post from the mailing list to make sure your collection names don't change.
  • Add support for generating scoped keys.
  • Make collection name encoding more robust. Make sure collection names are encoded identically for publishing events, running queries, and performing deletes.
  • Add support for grouping by multiple properties.
  • Add support for redirect URL creation.
  • Add support for HTTP and SOCKS proxies. Set KEEN_PROXY_URL to the proxy URL and KEEN_PROXY_TYPE to 'socks5' if you need to. These properties can also be set on the client instances as proxy_url and proxy_type.

  • Delegate the master_key fields from the Keen object.

  • Explicitly require CGI.
  • Use CGI.escape instead of URI.escape to get accurate URL encoding for certain characters
  • Add support for deletes (thanks again cbartlett!)
  • Allow event collection names for publishing/deleting methods to be symbols
  • Add batch publishing support
  • Allow event collection names for querying methods to be symbols. Thanks to cbartlett.
  • Fix support for non-https API URL testing
  • Allow configuration of the base API URL via the KEEN_API_URL environment variable. Useful for local testing and proxies.
  • BREAKING CHANGE! Added support for read and write scoped keys to reflect the new Keen IO security architecture. The advantage of scoped keys is finer grained permission control. Public clients that publish events (like a web browser) require a key that can write but not read. On the other hand, private dashboards and server-side querying processes require a Read key that should not be made public.
  • Improved logging and exception handling.
  • Added querying capabilities. A big thanks to ifeelgoods for contributing!
  • Removed API Key as a required field on Keen::Client. Only the Project ID is required to publish events.
  • You can continue to provide the API Key. Future features planned for this gem will require it. But for now, there is no keen-gem functionality that uses it.
  • Event collections are URI escaped to account for spaces.
  • User agent of API calls made more granular to aid in support cases.
  • Throw arguments error for nil event_collection and properties arguments.
  • Added beacon_url support
  • Add support for using em-synchrony with asynchronous calls

Questions & Support

For questions, bugs, or suggestions about this gem: File a Github Issue.

For other Keen-IO related technical questions: 'keen-io' on Stack Overflow

For general Keen IO discussion & feedback: 'keen-io-devs' Google Group


keen-gem is an open source project and we welcome your contributions. Fire away with issues and pull requests!

Running Tests

bundle exec rake spec - Run unit specs. HTTP is mocked.

bundle exec rake integration - Run integration specs with the real API. Requires env variables. See .travis.yml.

bundle exec rake synchrony - Run async publishing specs with EM::Synchrony.

Similarly, you can use guard to listen for changes to files and run specs.

bundle exec guard -g unit

bundle exec guard -g integration

bundle exec guard -g synchrony

Running a Local Console

You can spawn an irb session with the local files already loaded for debugging or experimentation.

$ bundle exec rake console
2.2.6 :001 > Keen
 => Keen 

Community Contributors

Thanks everyone, you rock!