The Python SDK lets you send events to Amplitude.
Install amplitude-analytics
with pip:
1pip install amplitude-analytics
Initialize the SDK before any events are instrumented. The API key for your Amplitude project is required. You can also pass a config object in this call. You can use the SDK client instance across requests after it's initialized.
1from amplitude import Amplitude2 3client = Amplitude(AMPLITUDE_API_KEY)
Name | Description | Default Value |
---|---|---|
api_key |
Required. String . The API key of the Amplitude project. Events sent by the client instance are in this project. Set when you initialize the client instance. |
None |
flush_queue_size |
Integer . Events wait in the buffer and are sent in a batch. The buffer is flushed when the number of events reaches flush_queue_size . |
200 |
flush_interval_millis |
Integer . Events wait in the buffer and are sent in a batch. The buffer is flushed every flush_interval_millis milliseconds. |
10 seconds |
flush_max_retries |
Integer . The number of times the client retries an event when the request returns an error. |
12 |
logger |
Logger. The logger instance used by Amplitude client. | python built-in logging: logging.getLogger(name) . |
min_id_length |
Integer . The minimum length of user_id and device_id . |
5 |
callback |
Function . Client level callback function. Takes three parameters:1. event: a Event instance 2. code: a integer of HTTP response code 3. message: a string message. |
None |
server_zone |
String . The server zone of the projects. Supports EU and US . For EU data residency, Change to EU . |
US |
server_url |
String . The API endpoint URL that events are sent to. Automatically selected by server_zone and use_batch . If this field is set with a string value instead of None , then server_zone and use_batch are ignored and the string value is used. |
https://api2.amplitude.com/2/httpapi |
use_batch |
Boolean . Whether to use batch API. By default, the SDK will use the default serverUrl . |
False |
storage_provider |
StorageProvider . Used to create storage instance to hold events in the storage buffer. Events in storage buffer are waiting to be sent. |
InMemoryStorageProvider |
opt_out |
Boolean . Opt out option. If set to True , client doesn't process and send events. |
False |
1def callback_func(event, code, message=None): 2 # callback function that takes three input parameters 3 # event: the event that triggered this callback 4 # code: status code of request response 5 # message: a optional string message for more detailed information 6 7client.configuration.api_key = "new api key" 8client.configuration.flush_max_retries = 5 9client.configuration.logger = logging.getLogger(__name__)10client.configuration.min_id_length = 711client.configuration.callback = callback_func12client.configuration.server_zone = "EU"13client.configuration.use_batch = True14client.configuration.server_url = "proxy url that forwarding the requests"15client.configuration.opt_out = False
To support high-performance environments, the SDK sends events in batches. Every event logged by track
method is queued in memory. Events are flushed in batches in background. You can customize batch behavior with flush_queue_size
and flush_interval_millis
. By default, the SDK is in regular mode with serverUrl
to https://api2.amplitude.com/2/httpapi
. For customers who want to send large batches of data at a time, switch to batch mode by setting use_batch
to true
to set setServerUrl to batch event upload API https://api2.amplitude.com/batch
. Both the regular mode and the batch mode use the same flush queue size and flush intervals.
1from amplitude import Amplitude 2 3client = Amplitude(AMPLITUDE_API_KEY) 4 5# Events queued in memory will flush when number of events exceed upload threshold 6# Default value is 200 7client.configuration.flush_queue_size = 100 8# Events queue will flush every certain milliseconds based on setting 9# Default value is 10 milliseconds10client.configuration.flush_interval_millis = 20000 # 20 seconds
Events represent how users interact with your application. For example, “Button Clicked” may be an action you want to note.
1from amplitude import Amplitude, BaseEvent 2 3client = Amplitude(AMPLITUDE_API_KEY) 4 5# Track a basic event 6# One of user_id and device_id is required 7event = BaseEvent(event_type="Button Clicked", user_id="User Id") 8client.track(event) 9 10# Track events with optional properties11client.track(12 BaseEvent(13 event_type="type of event",14 user_id="USER_ID",15 device_id="DEVICE_ID",16 event_properties={17 "source": "notification"18 }19))
User properties help you understand your users at the time they performed some action within your app such as their device details, their preferences, or language.
Identify is for setting the user properties of a particular user without sending any event. The SDK supports the operations set
, set_once
, unset
, add
, append
, prepend
, pre_insert
, post_insert
, and remove
on individual user properties. Declare the operations via a provided Identify interface. You can chain multiple operations together in a single Identify object. The Identify object is then passed to the Amplitude client to send to the server.
If the Identify call is sent after the event, the results of operations are visible immediately in the dashboard user’s profile area, but it won't appear in chart result until another event is sent after the Identify call. The identify call only affects events going forward. More details here.
The Identify object provides controls over setting user properties. An Identify object must first be instantiated, then Identify methods can be called on it, and finally the client makes a call with the Identify object.
1from amplitude import Identify, EventOptions2 3 4identify_obj=Identify()5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method sets the value of a user property. For example, you can set a role property of a user.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.set("location", "LAX")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method increments a user property by some numerical value. If the user property doesn't have a value set yet, it will be initialized to 0 before being incremented. For example, you can track a user's travel count.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.add("travel-count", 1)5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
You can use arrays as user properties. You can directly set arrays or use prepend
, append
, pre_insert
and post_insert
to generate an array.
This method prepends a value or values to a user property array. If the user property doesn't have a value set yet, it will be initialized to an empty list before the new values are prepended.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.prepend("visited-locations", "LAX")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method appends a value or values to a user property array. If the user property doesn't have a value set yet, it will be initialized to an empty list before the new values are appended.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.append("visited-locations", "SFO")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method pre-inserts a value or values to a user property if it doesn't exist in the user property yet. Pre-insert means inserting the values at the beginning of a given list. If the user property doesn't have a value set yet, it will be initialized to an empty list before the new values are pre-inserted. If the user property has an existing value, it will be no operation.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.pre_insert("unique-locations", "LAX")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method post-inserts a value or values to a user property if it doesn't exist in the user property yet. Post-insert means inserting the values at the end of a given list. If the user property doesn't have a value set yet, it will be initialized to an empty list before the new values are post-inserted. If the user property has an existing value, it will be no operation.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.post_insert("unique-locations", "SFO")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
This method removes a value or values to a user property if it exists in the user property. Remove means remove the existing value from the given list. If the item doesn't exist in the user property, it will be no operation.
1from amplitude import Identify, EventOptions2 3identify_obj=Identify()4identify_obj.remove("unique-locations", "JFK")5client.identify(identify_obj, EventOptions(user_id="USER_ID"))
Amplitude supports assigning users to groups and performing queries such as Count by Distinct on those groups. An example would be if you want to group your users based on what organization they're in by using an 'orgId'. You can designate Joe to be in 'orgId' '10' while Sue is in 'orgId' '15'. When performing a query in the Event Segmentation chart, you can then select "..performed by" 'orgId' to query the number of different organizations that have performed a specific event. As long as at least one member of that group has performed the specific event, that group is included in the count.
When setting groups, you need to define a group_type
and group_name
. In the previous example, 'orgId' is the group_type
and each of the values '10' and '15' are a group_name
. Another example of a group_type
could be 'sport' with group_name
values like 'tennis' and 'baseball'. You can use set_group()
to designate which groups a user belongs to. Note: This also sets the 'group_type:group_name' as a user property. This overwrites any existing group_name
value set for that user's group_type
, as well as the corresponding user property value. group_type
is a string and group_name
can be either a string or an array of strings to indicate a user being in multiple groups (for example, if Joe is in 'orgId' '10' and '16', then the group_name
would be '[10, 16]'). Here is what your code might look like.
1# set group with single group name2client.set_group(group_type="org_id", group_name="15",3 event_options=EventOptions(user_id="USER_ID"))4 5# set group with multiple group names6client.set_group(group_type="org_id", group_name=["15", "21"],7 event_options=EventOptions(user_id="USER_ID"))
Event level groups are set by groups
attribute of events
1# set groups when initial a event instance2event = BaseEvent("event_type", "user_id", groups={"org_id": ["15", "21"]})3 4# set groups for an existing instance5event["groups"] = {"sport": "soccer"}6 7client.track(event)
Use the Group Identify API to set or update the properties of particular groups. However, these updates will only affect events going forward.
The group_identify()
method accepts a group type and group name string parameter, as well as an Identify object that's applied to the group.
1identify_obj=Identify()2identify_obj.set("locale", "en-us")3client.group_identify(group_type="org-id", group_name="15", identify_obj=identify_obj)
The preferred method of tracking revenue for a user is to use revenue()
in conjunction with the provided Revenue interface. Revenue instances store each revenue transaction and allow you to define several special revenue properties, such as revenue_type
and product_id
, that are used in Amplitude's Event Segmentation and Revenue LTV charts. These Revenue instance objects are then passed into revenue
to send as revenue events to Amplitude. This allows Amplitude to automatically display data relevant to revenue in the platform. You can use this to track both in-app and non-in-app purchases.
To track revenue from a user, call revenue each time a user generates revenue. For example, 3 units of a product were purchased at $3.99.
1from amplitude import Revenue2 3revenue_obj = Revenue(price=3.99,4 quantity=3,5 product_id="com.company.productId")6client.revenue(revenue_obj, EventOptions(user_id="USER_ID"))
Name | Type | Description | Default |
---|---|---|---|
product_id (optional) |
string | An identifier for the product. Amplitude recommends something like the Google Play Store product ID. | null |
quantity (required) | int | The quantity of products purchased. revenue = quantity * price |
1 |
price (required) | Double | The price of the products purchased, and this can be negative. revenue = quantity * price |
null |
revenue_type (optional, required for revenue verification) |
String | The revenue type (for example, tax, refund, income). | null |
receipt (optional) | String | The receipt identifier of the revenue. | null |
receipt_sig (optional, required for revenue verification) |
String | The receipt signature of the revenue. | null |
properties (optional) | JSONObject | An object of event properties to include in the revenue event. | null |
The flush
method triggers the client to send buffered events.
1client.flush()
The add
method adds a plugin to Amplitude client instance. Plugins can help processing and sending events. Learn more about plugins..
1client.add(plugin_obj)
The remove
method removes the given plugin from the client instance if exists.
1client.remove(plugin_obj)
Use the shutdown
method to close the instance. A closed instance doesn't accept new events and tries to flush events left in the buffer. Then the client instance shuts down running threads.
In version v1.1.1 and higher, the shutdown
method is automatically registered to be called when the main thread exits.
1client.shutdown()
Plugins allow you to extend Amplitude SDK's behavior by, for example, modifying event properties (enrichment type) or sending to third-party APIs (destination type). A plugin is an object with methods setup()
and execute()
.
This method contains logic for preparing the plugin for use and has client
instance as a parameter. The expected return value is None
. A typical use for this method, is to copy configuration from client.configuration
or instantiate plugin dependencies. This method is called when you register the plugin to the client via client.add()
.
This method contains the logic for processing events and has event
instance as parameter. If used as enrichment type plugin, the expected return value is the modified/enriched event. If used as a destination type plugin, the expected return value is a map with keys: event
(BaseEvent), code
(number), and message
(string). This method is called for each event that's instrumented using the client interface, including Identify, GroupIdentify and Revenue events.
Here's an example of a plugin that modifies each event that's instrumented by adding an increment integer to event_id
property of an event.
1from threading import Lock 2from amplitude import Amplitude, EventPlugin, PluginType 3 4class AddEventIdPlugin(EventPlugin): 5 6 def __init__(self, start=0): 7 super().__init__(PluginType.ENRICHMENT) 8 self.current_id = start 9 self.configuration = None10 self.lock = Lock()11 12 def setup(self, client):13 self.configuration = client.configuration14 15 def execute(self, event):16 with self.lock:17 event.event_id = self.current_id18 self.current_id += 119 return event20 21client = Amplitude(AMPLITUDE_API_KEY)22client.add(AddInsertIdPlugin())
1from amplitude import Amplitude, EventPlugin, DestinationPlugin, PluginType 2import requests 3 4class MyDestinationPlugin(DestinationPlugin): 5 6 def __init__(self): 7 super().__init__() 8 # other init operations 9 self.url = "api endpoint url"10 self.configuration = None11 12 def setup(self, client):13 # setup plugin using client instance14 # triggered by client.add() method15 super().setup(client)16 self.configuration = client.configuration17 18 def execute(self, event):19 # process event using plugins in this destination plugin instance20 event = self.timeline.process(event)21 # send event to customized destination22 payload = '{"key":"secret", "event": ' + str(event) + '}'23 requests.post(self.url, data=payload)24 self.configuration.logger.info("Event sent")25 26 27client = Amplitude(AMPLITUDE_API_KEY)28client.add(MyDestinationPlugin())
Thanks for your feedback!
July 23rd, 2024
Need help? Contact Support
Visit Amplitude.com
Have a look at the Amplitude Blog
Learn more at Amplitude Academy
© 2024 Amplitude, Inc. All rights reserved. Amplitude is a registered trademark of Amplitude, Inc.