Experiment Python SDK

Official documentation for Amplitude Experiment's server-side Python SDK implementation.

This documentation has separate sections for remote and local evaluation.

Remote evaluation

Implements fetching variants for a user via remote evaluation.

Install

Install the Python Server SDK with pip.

1pip install amplitude-experiment

 1from amplitude_experiment import Experiment, RemoteEvaluationConfig, RemoteEvaluationClient, User
 2 
 3# (1) Initialize the experiment client
 4experiment = Experiment.initialize_remote('<DEPLOYMENT_KEY>')
 5 
 6# (2) Fetch variants for a user
 7user = User(
 8    device_id="abcdefg",
 9    user_id="user@company.com",
10    user_properties={
11        'premium': True
12    }
13)
14variants = experiment.fetch_v2(user)
15 
16# (3) Access a flag's variant
17variant = variants['YOUR-FLAG-KEY']
18if variant:
19    if variant.value == 'on':
20        # Flag is on
21    else:
22        # Flag is off

Initialize

The SDK client should be initialized in your server on startup. The deployment key argument passed into the api_key parameter must live within the same project that you are sending analytics events to.

1Experiment.initialize_remote(api_key, config = None) : RemoteEvaluationClient

1experiment = Experiment.initialize_remote('<DEPLOYMENT_KEY>', Config())

Configuration

You can configure the SDK client on initialization.

Fetch

Fetches variants for a user and returns the results. This function remote evaluates the user for flags associated with the deployment used to initialize the SDK client.

1fetch_v2(user: User) : Variants

1user = User(
2    device_id="abcdefg",
3    user_id="user@company.com",
4    user_properties={
5        'premium': True
6    }
7)
8variants = experiment.fetch_v2(user)

After fetching variants for a user, you may to access the variant for a specific flag.

1variant = variants['YOUR-FLAG-KEY']
2if variant:
3    if variant.value == 'on':
4        # Flag is on
5    else:
6        # Flag is off

Fetch async

The fetch method is synchronous. To fetch asynchronously, you can use fetch_async method

1fetch_async_v2(user: User, callback)

1def fetch_callback(user, variants):
2  variant = variants['YOUR-FLAG-KEY']
3  if variant:
4    if variant.value == 'on':
5      # Flag is on
6    else:
7      # Flag is off
8 
9experiment.fetch_async_v2(user, fetch_callback)

Local evaluation

Implements evaluating variants for a user via local evaluation. If you plan on using local evaluation, you should understand the tradeoffs.

Install

Install the Python Server SDK's local evaluation.

Install the Python Server SDK with pip.

1pip install amplitude-experiment

 1# (1) Initialize the local evaluation client with a server deployment key.
 2experiment = Experiment.initialize_local("DEPLOYMENT_KEY", LocalEvaluationConfig(
 3  # (Recommended) Enable local evaluation cohort targeting.
 4  cohort_sync_config=CohortSyncConfig(api_key="API_KEY", secret_key="SECRET_KEY")
 5))
 6 
 7# (2) Start the local evaluation client.
 8experiment.start()
 9 
10# (3) Evaluate a user.
11user = User(
12    device_id="abcdefg",
13    user_id="user@company.com",
14    user_properties={
15        'premium': True
16    }
17)
18variants = experiment.evaluate_v2(user)

Initialize

Initializes a local evaluation client.

1Experiment.initialize_local(api_key, config = None) : LocalEvaluationClient

Configuration

You can configure the SDK client on initialization.

LocalEvaluationConfig

AssignmentConfig

CohortSyncConfig

Start

Start the local evaluation client, pre-fetching local evaluation mode flag configs for evaluation and starting the flag config poller at the configured interval.

1start()

You should await the result of start() to ensure that flag configs are ready to be used before calling evaluate_v2()

1experiment.start()

Evaluate

Executes the evaluation logic using the flags pre-fetched on start(). Evaluate must be given a user object argument and can optionally be passed an array of flag keys if only a specific subset of required flag variants are required.

1evaluate_v2(self, user: User, flag_keys: List[str]) : Dict[str, Variant]

 1# The user to evaluate
 2user = User(user_id='test_user')
 3 
 4# Evaluate all flag variants
 5all_variants = experiment.evaluate_v2(user)
 6 
 7# Evaluate a specific subset of flag variants
 8specific_variants = experiment.evaluate_v2(user, ["<FLAG_KEY_1>", "<FLAG_KEY_2>"])
 9 
10# Access a variant
11variant = all_variants["<FLAG_KEY>"]
12if variant.value == 'on':
13    # Flag is on
14else:
15    # Flag is off

Local evaluation cohort targeting

Since version 1.4.0, the local evaluation SDK client supports downloading cohorts for local evaluation targeting. You must configure the cohort_sync_config option with the analytics api_key and secret_key on initialization to enable this support.

1experiment = Experiment.initialize_local("DEPLOYMENT_KEY", LocalEvaluationConfig(
2  # (Recommended) Enable local evaluation cohort targeting.
3  cohort_sync_config=CohortSyncConfig(api_key="API_KEY", secret_key="SECRET_KEY")
4))

Access Amplitude cookies

If you're using the Amplitude Analytics SDK on the client-side, the Python server SDK provides an AmplitudeCookie class with convenience functions for parsing and interacting with the Amplitude identity cookie. This is useful for ensuring that the Device ID on the server matches the Device ID set on the client, especially if the client hasn't yet generated a Device ID.

 1import uuid
 2from amplitude_experiment import AmplitudeCookie
 3 
 4# grab amp device id if present
 5amp_cookie_name = AmplitudeCookie.cookie_name('amplitude-api-key')
 6device_id = nil
 7if request.cookies.get(amp_cookie_name):
 8  device_id = AmplitudeCookie.parse(request.cookies.get(amp_cookie_name)).device_id
 9 
10if device_id is None:
11  # deviceId doesn't exist, set the Amplitude Cookie
12  device_id = str(uuid.uuid4())
13  amp_cookie_value = AmplitudeCookie.generate(device_id)
14  resp.set_cookie(amp_cookie_name, {
15    "value": amp_cookie_value,
16    "domain": ".your-domain.com",  # this should be the same domain used by the Amplitude JS SDK
17    "httponly": False,
18    "secure": False
19  })