Official Python client for the Keen IO API. Build analytics features directly into your Python apps.
Latest commit efbed79 May 22, 2017 @josephwegner josephwegner committed on GitHub Merge pull request #127 from keenlabs/jm_0.4.0
Bump version to 0.4.0.


Keen IO Official Python Client Library

Build status

This is the official Python Client for the Keen IO API. The Keen IO API lets developers build analytics features directly into their apps.

This is still under active development. Stay tuned for improvements!


Use pip to install!

pip install keen

This client is known to work on Python 2.6, 2.7, 3.2, 3.3, 3.4, 3.5 and 3.6.

For versions of Python < 2.7.9, you’ll need to install pyasn1, ndg-httpsclient, pyOpenSSL.


To use this client with the Keen IO API, you have to configure your Keen IO Project ID and its access keys (if you need an 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. The recommended way to set this configuration information is via the environment. The keys you can set are KEEN_PROJECT_ID, KEEN_WRITE_KEY, KEEN_READ_KEY, and KEEN_MASTER_KEY. As per the Principle of Least Privilege, it's recommended that you not use the master_key if not necessary. This SDK will expect and use the precise key for a given operation, and throw an exception in cases of misuse.

If you don't want to use environment variables for some reason, you can directly set values as follows:

keen.project_id = "xxxx"
keen.write_key = "yyyy"
keen.read_key = "zzzz"
keen.master_key = "abcd" # not required for typical usage

For information on how to configure unique client instances, take a look at the Advanced Usage section below.

Send Events to Keen IO

Once you've set KEEN_PROJECT_ID and KEEN_WRITE_KEY, sending events is simple:

keen.add_event("sign_ups", {
    "username": "lloyd",
    "referred_by": "harry"

Send Batch Events to Keen IO

You can upload Events in a batch, like so:

# uploads 4 events total - 2 to the "sign_ups" collection and 2 to the "purchases" collection
    "sign_ups": [
        { "username": "nameuser1" },
        { "username": "nameuser2" }
    "purchases": [
        { "price": 5 },
        { "price": 6 }

That's it! After running your code, check your Keen IO Project to see the event/events has been added.

Do analysis with Keen IO

Here are some examples of querying. Let's assume you've added some events to the "purchases" collection. For more code samples, take a look at Keen's docs

keen.count("purchases", timeframe="this_14_days") # => 100
keen.sum("purchases", target_property="price", timeframe="this_14_days") # => 10000
keen.minimum("purchases", target_property="price", timeframe="this_14_days") # => 20
keen.maximum("purchases", target_property="price", timeframe="this_14_days") # => 100
keen.average("purchases", target_property="price", timeframe="this_14_days") # => 49.2

keen.sum("purchases", target_property="price", group_by="", timeframe="this_14_days") # => [{ "": 123, "result": 240 }, { ... }]

keen.count_unique("purchases", target_property="", timeframe="this_14_days") # => 3
keen.select_unique("purchases", target_property="", timeframe="this_14_days") # => ["", ""]

keen.extraction("purchases", timeframe="today") # => [{ "price" => 20, ... }, { ... }]

            "analysis_type": "sum",
            "target_property": "price"
            "analysis_type": "average",
            "target_property": "price"
) # => {"total":10329.03, "average":933.93}

step1 = {
    "event_collection": "sign_ups",
    "actor_property": ""
step2 = {
    "event_collection": "purchases",
    "actor_property": ""
keen.funnel([step1, step2], timeframe="today") # => [2039, 201]

To return the full API response from a funnel analysis (as opposed to the singular "result" key), set all_keys=True.

For example, keen.funnel([step1, step2], timeframe="today", all_keys=True) would return "result", "actors" and "steps" keys.

Delete Events

The Keen IO API allows you to delete events from event collections, optionally supplying filters, timeframe or timezone to narrow the scope of what you would like to delete.

You'll need to set your master_key.

keen.delete_events("event_collection", filters=[{"property_name": 'username', "operator": 'eq', "property_value": 'Bob'}])

Advanced Usage

See below for more options.

Check Batch Upload Response For Errors

When you upload events in a batch, some of them may succeed and some of them may have errors. The Keen API returns information on each. Here's an example:

Upload code (remember, Keen IO doesn't allow periods in property names):

response = keen.add_events({
    "sign_ups": [
        { "username": "nameuser1" },
        { "username": "nameuser2", "": 1 }
    "purchases": [
        { "price": 5 },
        { "price": 6 }

That code would result in the following API JSON response:

    "sign_ups": [
        {"success": true},
        {"success": false, "error": {"name": "some_error_name", "description": "some longer description"}}
    "purchases": [
        {"success": true},
        {"success": true}

So in python, to check on the results of your batch, you'd have code like so:

batch = {
    "sign_ups": [
        { "username": "nameuser1" },
        { "username": "nameuser2", "": 1 }
    "purchases": [
        { "price": 5 },
        { "price": 6 }
response = keen.add_events(batch)

for collection in response:
    collection_result = response[collection]
    event_count = 0
    for individual_result in collection_result:
        if not individual_result["success"]:
            print("Event had error! Collection: '{}'. Event body: '{}'.".format(collection, batch[collection][event_count]))
        event_count += 1

Configure Unique Client Instances

If you intend to send events or query from different projects within the same python file, you'll need to set up unique client instances (one per project). You can do this by assigning an instance of KeenClient to a variable like so:

from keen.client import KeenClient

client = KeenClient(
    project_id="xxxx",  # your project ID for collecting cycling data
    master_key="abcd" # not required for typical usage

client_hike = KeenClient(
    project_id="xxxx",  # your project ID for collecting hiking data (different from the one above)
    master_key="abcd" # not required for typical usage

You can send events like this:

# add an event to an event collection in your cycling project

# or add an event to an event collection in your hiking project

Similarly, you can query events like this:


Saved Queries

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

# Create your KeenClient
from keen.client import KeenClient

client = KeenClient(
    project_id="xxxx",  # your project ID
    master_key="abcd" # Most Saved Query functionality requires master_key

# Create a saved query
saved_query_attributes = {
    # NOTE : For now, refresh_rate must explicitly be set to 0 unless you
    # intend to create a Cached Query.
    "refresh_rate": 0,
    "query": {
        "analysis_type": "count",
        "event_collection": "purchases",
        "timeframe": "this_2_weeks",
        "filters": [{
            "property_name": "price",
            "operator": "gte",
            "property_value": 1.00

client.saved_queries.create("saved-query-name", saved_query_attributes)

# Get all saved queries

# Get one saved query

# Get saved query with results

# NOTE : Updating Saved Queries requires sending the entire query definition. Any attribute not
# sent is interpreted as being cleared/removed. This means that properties set via another
# client, including the Projects Explorer Web UI, will be lost this way.
# The update() function makes this easier by allowing client code to just specify the
# properties that need updating. To do this, it will retrieve the existing query definition
# first, which means there will be two HTTP requests. Use update_full() in code that already
# has a full query definition that can reasonably be expected to be current.

# Update a saved query to now be a cached query with the minimum refresh rate of 4 hrs...

# ...using partial update:
client.saved_queries.update("saved-query-name", { "refresh_rate": 14400 })

# ...using full update, if we've already fetched the query definition:
saved_query_attributes["refresh_rate"] = 14400
client.saved_queries.update_full("saved-query-name", saved_query_attributes)

# Update a saved query to a new resource name...

# ...using partial update:
client.saved_queries.update("saved-query-name", { "query_name": "cached-query-name" })

# ...using full update, if we've already fetched the query definition or have it lying around
# for whatever reason. We send "refresh_rate" again, along with the entire definition, or else
# it would be reset:
saved_query_attributes["query_name"] = "cached-query-name"
client.saved_queries.update_full("saved-query-name", saved_query_attributes)

# Delete a saved query (use the new resource name since we just changed it)

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.add_event("sign_ups", {
    "keen": {
        "timestamp": "2012-07-06T02:09:10.141Z"
    "username": "lloyd",
    "referred_by": "harry"

Get from Keen IO with a Timeout

By default, GET requests will timeout after 305 seconds. If you want to manually override this, you can create a KeenClient with the "get_timeout" parameter. This client will fail GETs if no bytes have been returned by the server in the specified time. For example:

from keen.client import KeenClient

client = KeenClient(

This will cause queries such as count(), sum(), and average() to timeout after 100 seconds. If this timeout limit is hit, a requests.Timeout will be raised. Due to a bug in the requests library, you might also see an SSLError (#1294)

Send to Keen IO with a Timeout

By default, POST requests will timeout after 305 seconds. If you want to manually override this, you can create a KeenClient with the "post_timeout" parameter. This client will fail POSTs if no bytes have been returned by the server in the specified time. For example:

from keen.client import KeenClient

client = KeenClient(

This will cause both add_event() and add_events() to timeout after 100 seconds. If this timeout limit is hit, a requests.Timeout will be raised. Due to a bug in the requests library, you might also see an SSLError (

Create Scoped Keys

The Python client enables you to create Scoped Keys easily. For example:

from keen.client import KeenClient
from keen import scoped_keys


write_key = scoped_keys.encrypt(api_key, {"allowed_operations": ["write"]})
read_key = scoped_keys.encrypt(api_key, {"allowed_operations": ["read"]})

write_key and read_key now contain scoped keys based on your master API key.


To run tests:

python test



  • SavedQueriesInterface.create() now accepts a dict as the query definition.
  • get_collection() and get_all_collections() now only require a Read Key instead of Master.
  • SavedQueriesInterface.update() now performs partial updates. update_full() exhibits old behavior.
  • Misc documentation updates.


  • Fix broken releases.


  • Add Keen-Sdk header to all requests, containing the SDK version.


  • Fix incorrect README.


  • Return JSON response when uploading events in a batch.


  • Removed unused Padding from requirements.txt to make python 3.x installs cleaner.


  • Replaced defunct pycrypto library with cryptodome.
  • Fixed UnicodeDecodeError under PY3 while installing in Windows.


  • Updated documentation


  • Added status code to JSON parse error response


  • Added support for python 3.5


  • Fixed bug with scoped key generation not working with newer Keen projects.


  • Added saved_queries support
  • Added Python 3.4 support


  • Added base_url as a possible env variable


  • Updated error handling to except ValueError


  • Fixed timestamp overriding keen addons
  • Added get_collection and get_all_collections methods


  • Added all_keys parameter which allows users to expose all keys in query response.
  • Added delete_events method.


  • Added better error handling to surface all errors from HTTP API calls.


  • Added compatibility for pip 1.0


  • Added compatibility for pip < 1.5.6


  • Made requirements more flexible.


  • Added requirements.txt to pypi package.


  • Fixed requirements in
  • Updated test inputs and documentation.


  • Added `master_key` parameter.


  • Mocked tests.
  • Added `median` query method.
  • Added support for $python test.


  • Upgraded to requests==2.5.1


  • Added `max_age` parameter for caching.


  • Added client configurable timeout to gets.


  • Added `percentile` query method.


  • Support `interval` parameter for multi analyses on the keen module.


  • Reuse internal requests' session inside an instance of KeenApi.


  • Support `property_names` parameter for extractions.


  • Added client configurable timeout to posts.
  • Upgraded to requests==2.2.1.


  • Fixed sys.version_info issue with Python 2.6.


  • Added interval to multi_analysis.


  • Added stacktrace_id and unique_id to Keen API errors.


  • Added add_events method to keen/ so it can be used at a module level.
  • Added method to generate image beacon URLs.


  • Added support for publishing events in batches
  • Added support for configuring client automatically from environment
  • Added methods on keen module directly


  • Added querying support


  • Bugfix to use write key when sending events - do not use 0.1.6!


  • Changed project token -> project ID.
  • Added support for read and write scoped keys.
  • Added support for generating scoped keys yourself.
  • Added support for python 2.6, 3.2, and 3.3


  • Added documentation.

To Do

  • Asynchronous insert
  • Scoped keys

Questions & Support

If you have any questions, bugs, or suggestions, please report them via Github Issues. We'd love to hear your feedback and ideas!


This is an open source project and we love involvement from the community! Hit us up with pull requests and issues.