SparkPost
Anymail integrates with the Bird SparkPost email service, using their Transmissions API.
Changed in version 8.0: Earlier Anymail versions used the official Python sparkpost API client.
That library is no longer maintained, and Anymail now calls SparkPost’s HTTP API
directly. This change should not affect most users, but you should make sure you
provide SPARKPOST_API_KEY
in your
Anymail settings (Anymail doesn’t check environment variables), and if you are
using Anymail’s esp_extra you will need to update that
to use Transmissions API parameters.
Settings
EMAIL_BACKEND
To use Anymail’s SparkPost backend, set:
EMAIL_BACKEND = "anymail.backends.sparkpost.EmailBackend"
in your settings.py.
SPARKPOST_API_KEY
A SparkPost API key with at least the “Transmissions: Read/Write” permission. (Manage API keys in your SparkPost account API keys.)
ANYMAIL = { ... "SPARKPOST_API_KEY": "<your API key>", }
Anymail will also look for SPARKPOST_API_KEY
at the
root of the settings file if neither ANYMAIL["SPARKPOST_API_KEY"]
nor ANYMAIL_SPARKPOST_API_KEY
is set.
Changed in version 8.0: This setting is required. If you store your API key in an environment variable, load
it into your Anymail settings: "SPARKPOST_API_KEY": os.environ["SPARKPOST_API_KEY"]
.
(Earlier Anymail releases used the SparkPost Python library, which would look for
the environment variable.)
SPARKPOST_SUBACCOUNT
Added in version 8.0.
An optional SparkPost subaccount numeric id. This can be used, along with the API key for the master account, to send mail on behalf of a subaccount. (Do not set this when using a subaccount’s own API key.)
Like all Anymail settings, you can include this in the global settings.py ANYMAIL dict
to apply to all sends, or supply it as a get_connection()
keyword parameter (connection = get_connection(subaccount=123)
) to send a particular
message with a subaccount. See Mixing email backends for more information on using
connections.
SPARKPOST_API_URL
The SparkPost API Endpoint to use. The default is "https://api.sparkpost.com/api/v1"
.
Set this to use a SparkPost EU account, or to work with any other API endpoint including SparkPost Enterprise API and SparkPost Labs.
ANYMAIL = { ... "SPARKPOST_API_URL": "https://api.eu.sparkpost.com/api/v1", # use SparkPost EU }
You must specify the full, versioned API endpoint as shown above (not just the base_uri).
SPARKPOST_TRACK_INITIAL_OPEN_AS_OPENED
Added in version 8.1.
Boolean, default False
. When using Anymail’s tracking webhooks, whether to report
SparkPost’s “Initial Open” event as an Anymail normalized “opened” event.
(SparkPost’s “Open” event is always normalized to Anymail’s “opened” event.
See Status tracking webhooks below.)
esp_extra support
To use SparkPost features not directly supported by Anymail, you can set
a message’s esp_extra
to a dict
of transmissions API request body data. Anymail will deeply merge your overrides
into the normal API payload it has constructed, with esp_extra taking precedence
in conflicts.
Example (you probably wouldn’t combine all of these options at once):
message.esp_extra = { "options": { # Treat as transactional for unsubscribe and suppression: "transactional": True, # Override your default dedicated IP pool: "ip_pool": "transactional_pool", }, # Add a description: "description": "Test-run for new templates", "content": { # Use draft rather than published template: "use_draft_template": True, # Use an A/B test: "ab_test_id": "highlight_support_links", }, # Use a stored recipients list (overrides message to/cc/bcc): "recipients": { "list_id": "design_team" }, }
Note that including "recipients"
in esp_extra will completely override the
recipients list Anymail generates from your message’s to/cc/bcc fields, along with any
per-recipient merge_data
and
merge_metadata
.
(You can also set "esp_extra"
in Anymail’s global send defaults
to apply it to all messages.)
Limitations and quirks
- Anymail’s `message_id` is SparkPost’s `transmission_id`
The
message_id
Anymail sets on a message’sanymail_status
and in normalized webhookAnymailTrackingEvent
data is actually what SparkPost calls “transmission_id”.Like Anymail’s message_id for other ESPs, SparkPost’s transmission_id (together with the recipient email address), uniquely identifies a particular message instance in tracking events.
(The transmission_id is the only unique identifier available when you send your message. SparkPost also has something called “message_id”, but that doesn’t get assigned until after the send API call has completed.)
If you are working exclusively with Anymail’s normalized message status and webhook events, the distinction won’t matter: you can consistently use Anymail’s
message_id
. But if you are also working with raw webhook esp_event data or SparkPost’s events API, be sure to think “transmission_id” wherever you’re speaking to SparkPost.- Single tag
Anymail uses SparkPost’s “campaign_id” to implement message tagging. SparkPost only allows a single campaign_id per message. If your message has two or more
tags
, you’ll get anAnymailUnsupportedFeature
error—or if you’ve enabledANYMAIL_IGNORE_UNSUPPORTED_FEATURES
, Anymail will use only the first tag.(SparkPost’s “recipient tags” are not available for tagging messages. They’re associated with individual addresses in stored recipient lists.)
- AMP for Email
SparkPost supports sending AMPHTML email content. To include it, use
message.attach_alternative("...AMPHTML content...", "text/x-amp-html")
(and be sure to also include regular HTML and/or text bodies, too).Added in version 8.0.
- Extra header limitations
SparkPost’s API silently ignores certain email headers (specified via Django’s headers or extra_headers or Anymail’s
merge_headers
). In particular, attempts to provide a custom List-Unsubscribe header will not work; the message will be sent with SparkPost’s own subscription management headers. (The list of allowed custom headers does not seem to be documented.)
- Features incompatible with template_id
When sending with a
template_id
, SparkPost doesn’t support attachments, inline images, extra headers,reply_to
,cc
recipients, or overriding thefrom_email
,subject
, or body (text or html) when sending the message. Some of these can be defined in the template itself, but SparkPost (often) silently drops them when supplied to their Transmissions send API.Changed in version 11.0: Using features incompatible with
template_id
will raise anAnymailUnsupportedFeature
error. In earlier releases, Anymail would pass the incompatible content to SparkPost’s API, which in many cases would silently ignore it and send the message anyway.These limitations only apply when using stored templates (with a template_id), not when using SparkPost’s template language for on-the-fly templating in a message’s subject, body, etc.
- Envelope sender may use domain only
Anymail’s
envelope_sender
is used to populate SparkPost’s'return_path'
parameter. Anymail supplies the full email address, but depending on your SparkPost configuration, SparkPost may use only the domain portion and substitute its own encoded mailbox before the @.- Multiple from_email addresses
Prior to November, 2020, SparkPost supporting sending messages with multiple From addresses. (This is technically allowed by email specs, but many ISPs bounce such messages.) Anymail v8.1 and earlier will pass multiple
from_email
addresses to SparkPost’s API.SparkPost has since dropped support for more than one from address, and now issues error code 7001 “No sending domain specified”. To avoid confusion, Anymail v8.2 treats multiple from addresses as an unsupported feature in the SparkPost backend.
Changed in version 8.2.
Batch sending/merge and ESP templates
SparkPost offers both ESP stored templates and batch sending with per-recipient merge data.
You can use a SparkPost stored template by setting a message’s
template_id
to the
template’s unique id. (When using a stored template, SparkPost prohibits
setting the EmailMessage’s subject, text body, or html body, and has
several other limitations.)
Alternatively, you can refer to merge fields directly in an EmailMessage’s subject, body, and other fields—the message itself is used as an on-the-fly template.
In either case, supply the merge data values with Anymail’s
normalized merge_data
and merge_global_data
message attributes.
message = EmailMessage( ... to=["[email protected]", "Bob <[email protected]>"] ) message.template_id = "11806290401558530" # SparkPost id message.from_email = None # must set after constructor (see below) message.merge_data = { '[email protected]': {'name': "Alice", 'order_no': "12345"}, '[email protected]': {'name': "Bob", 'order_no': "54321"}, } message.merge_global_data = { 'ship_date': "May 15", # Can use SparkPost's special "dynamic" keys for nested substitutions (see notes): 'dynamic_html': { 'status_html': "<a href='https://example.com/order/{{order_no}}'>Status</a>", }, 'dynamic_plain': { 'status_plain': "Status: https://example.com/order/{{order_no}}", }, }
When using a template_id
, you must set the
message’s from_email
to None
as shown above. SparkPost does not permit
specifying the from address at send time when using a stored template.
See SparkPost’s substitutions reference for more information on templates and
batch send with SparkPost. If you need the special “dynamic” keys for nested substitutions,
provide them in Anymail’s merge_global_data
as shown in the example above. And if you want use_draft_template
behavior, specify that
in esp_extra.
Status tracking webhooks
If you are using Anymail’s normalized status tracking, set up the webhook in your SparkPost configuration under “Webhooks”:
Target URL:
https://yoursite.example.com/anymail/sparkpost/tracking/
Authentication: choose “Basic Auth.” For username and password enter the two halves of the random:random shared secret you created for your
ANYMAIL_WEBHOOK_SECRET
Django setting. (Anymail doesn’t support OAuth webhook auth.)Events: you can leave “All events” selected, or choose “Select individual events” to pick the specific events you’re interested in tracking.
SparkPost will report these Anymail event_type
s:
queued, rejected, bounced, deferred, delivered, opened, clicked, complained, unsubscribed,
subscribed.
By default, Anymail reports SparkPost’s “Open”—but not its “Initial Open”—event
as Anymail’s normalized “opened” event_type
.
This avoids duplicate “opened” events when both SparkPost types are enabled.
Added in version 8.1: To receive SparkPost “Initial Open” events as Anymail’s “opened”, set
"SPARKPOST_TRACK_INITIAL_OPEN_AS_OPENED": True
in your ANYMAIL settings dict. You will probably want to disable SparkPost “Open”
events when using this setting.
Changed in version 8.1: SparkPost’s “AMP Click” and “AMP Open” are reported as Anymail’s “clicked” and “opened” events. If you enable the SPARKPOST_TRACK_INITIAL_OPEN_AS_OPENED setting, “AMP Initial Open” will also map to “opened.” (Earlier Anymail releases reported all AMP events as “unknown”.)
The event’s esp_event
field will be
a single, raw SparkPost event. (Although SparkPost calls webhooks with batches of events,
Anymail will invoke your signal receiver separately for each event in the batch.)
The esp_event is the raw, wrapped json event structure as provided by SparkPost:
{'msys': {'<event_category>': {...<actual event data>...}}}
.
Inbound webhook
If you want to receive email from SparkPost through Anymail’s normalized inbound handling, follow SparkPost’s Enabling Inbound Email Relaying guide to set up Anymail’s inbound webhook.
The target parameter for the Relay Webhook will be:
https://random:random@yoursite.example.com/anymail/sparkpost/inbound/
random:random is an
ANYMAIL_WEBHOOK_SECRET
shared secretyoursite.example.com is your Django site