Top Level API

This is the user facing API of the SDK. It’s exposed as sentry_sdk. With this API you can implement a custom performance monitoring or error reporting solution.

Initializing the SDK

class sentry_sdk.client.ClientConstructor(dsn=None, *, max_breadcrumbs=100, release=None, environment=None, server_name=None, shutdown_timeout=2, integrations=[], in_app_include=[], in_app_exclude=[], default_integrations=True, dist=None, transport=None, transport_queue_size=100, sample_rate=1.0, send_default_pii=None, http_proxy=None, https_proxy=None, ignore_errors=[], max_request_body_size='medium', socket_options=None, keep_alive=None, before_send=None, before_breadcrumb=None, debug=None, attach_stacktrace=False, ca_certs=None, propagate_traces=True, traces_sample_rate=None, traces_sampler=None, profiles_sample_rate=None, profiles_sampler=None, profiler_mode=None, profile_lifecycle='manual', profile_session_sample_rate=None, auto_enabling_integrations=True, disabled_integrations=None, auto_session_tracking=True, send_client_reports=True, _experiments={}, proxy_headers=None, instrumenter='sentry', before_send_transaction=None, project_root=None, enable_tracing=None, include_local_variables=True, include_source_context=True, trace_propagation_targets=['.*'], functions_to_trace=[], event_scrubber=None, max_value_length=1024, enable_backpressure_handling=True, error_sampler=None, enable_db_query_source=True, db_query_source_threshold_ms=100, spotlight=None, cert_file=None, key_file=None, custom_repr=None, add_full_stack=False, max_stack_frames=100)[source]
__init__(dsn=None, *, max_breadcrumbs=100, release=None, environment=None, server_name=None, shutdown_timeout=2, integrations=[], in_app_include=[], in_app_exclude=[], default_integrations=True, dist=None, transport=None, transport_queue_size=100, sample_rate=1.0, send_default_pii=None, http_proxy=None, https_proxy=None, ignore_errors=[], max_request_body_size='medium', socket_options=None, keep_alive=None, before_send=None, before_breadcrumb=None, debug=None, attach_stacktrace=False, ca_certs=None, propagate_traces=True, traces_sample_rate=None, traces_sampler=None, profiles_sample_rate=None, profiles_sampler=None, profiler_mode=None, profile_lifecycle='manual', profile_session_sample_rate=None, auto_enabling_integrations=True, disabled_integrations=None, auto_session_tracking=True, send_client_reports=True, _experiments={}, proxy_headers=None, instrumenter='sentry', before_send_transaction=None, project_root=None, enable_tracing=None, include_local_variables=True, include_source_context=True, trace_propagation_targets=['.*'], functions_to_trace=[], event_scrubber=None, max_value_length=1024, enable_backpressure_handling=True, error_sampler=None, enable_db_query_source=True, db_query_source_threshold_ms=100, spotlight=None, cert_file=None, key_file=None, custom_repr=None, add_full_stack=False, max_stack_frames=100)[source]

Initialize the Sentry SDK with the given parameters. All parameters described here can be used in a call to sentry_sdk.init().

Parameters:

  • dsn (Optional[str]) –

    The DSN tells the SDK where to send the events.

    If this option is not set, the SDK will just not send any data.

    The dsn config option takes precedence over the environment variable.

    Learn more about DSN utilization.

  • debug (Optional[bool]) –

    Turns debug mode on or off.

    When True, the SDK will attempt to print out debugging information. This can be useful if something goes wrong with event sending.

    The default is always False. It’s generally not recommended to turn it on in production because of the increase in log output.

    The debug config option takes precedence over the environment variable.

  • release (Optional[str]) –

    Sets the release.

    If not set, the SDK will try to automatically configure a release out of the box but it’s a better idea to manually set it to guarantee that the release is in sync with your deploy integrations.

    Release names are strings, but some formats are detected by Sentry and might be rendered differently.

    See the releases documentation to learn how the SDK tries to automatically configure a release.

    The release config option takes precedence over the environment variable.

    Learn more about how to send release data so Sentry can tell you about regressions between releases and identify the potential source in the product documentation.

  • environment (Optional[str]) –

    Sets the environment. This string is freeform and set to production by default.

    A release can be associated with more than one environment to separate them in the UI (think staging vs production or similar).

    The environment config option takes precedence over the environment variable.

  • dist (Optional[str]) –

    The distribution of the application.

    Distributions are used to disambiguate build or deployment variants of the same release of an application.

    The dist can be for example a build number.

  • sample_rate (float) –

    Configures the sample rate for error events, in the range of 0.0 to 1.0.

    The default is 1.0, which means that 100% of error events will be sent. If set to 0.1, only 10% of error events will be sent.

    Events are picked randomly.

  • error_sampler (Optional[Callable[[Event, Dict[str, Any]], Union[float, bool]]]) –

    Dynamically configures the sample rate for error events on a per-event basis.

    This configuration option accepts a function, which takes two parameters (the event and the hint), and which returns a boolean (indicating whether the event should be sent to Sentry) or a floating-point number between 0.0 and 1.0, inclusive.

    The number indicates the probability the event is sent to Sentry; the SDK will randomly decide whether to send the event with the given probability.

    If this configuration option is specified, the sample_rate option is ignored.

  • ignore_errors (Sequence[Union[type, str]]) –

    A list of exception class names that shouldn’t be sent to Sentry.

    Errors that are an instance of these exceptions or a subclass of them, will be filtered out before they’re sent to Sentry.

    By default, all errors are sent.

  • max_breadcrumbs (int) –

    This variable controls the total amount of breadcrumbs that should be captured.

    This defaults to 100, but you can set this to any number.

    However, you should be aware that Sentry has a maximum payload size and any events exceeding that payload size will be dropped.

  • attach_stacktrace (bool) –

    When enabled, stack traces are automatically attached to all messages logged.

    Stack traces are always attached to exceptions; however, when this option is set, stack traces are also sent with messages.

    This option means that stack traces appear next to all log messages.

    Grouping in Sentry is different for events with stack traces and without. As a result, you will get new groups as you enable or disable this flag for certain events.

  • send_default_pii (Optional[bool]) –

    If this flag is enabled, certain personally identifiable information (PII) is added by active integrations.

    If you enable this option, be sure to manually remove what you don’t want to send using our features for managing Sensitive Data.

  • event_scrubber (Optional[EventScrubber]) –

    Scrubs the event payload for sensitive information such as cookies, sessions, and passwords from a denylist.

    It can additionally be used to scrub from another pii_denylist if send_default_pii is disabled.

    See how to configure the scrubber here.

  • include_source_context (Optional[bool]) –

    When enabled, source context will be included in events sent to Sentry.

    This source context includes the five lines of code above and below the line of code where an error happened.

  • include_local_variables (Optional[bool]) – When enabled, the SDK will capture a snapshot of local variables to send with the event to help with debugging.

  • add_full_stack (bool) –

    When capturing errors, Sentry stack traces typically only include frames that start the moment an error occurs.

    But if the add_full_stack option is enabled (set to True), all frames from the start of execution will be included in the stack trace sent to Sentry.

  • max_stack_frames (Optional[int]) – This option limits the number of stack frames that will be captured when add_full_stack is enabled.

  • server_name (Optional[str]) –

    This option can be used to supply a server name.

    When provided, the name of the server is sent along and persisted in the event.

    For many integrations, the server name actually corresponds to the device hostname, even in situations where the machine is not actually a server.

  • project_root (Optional[str]) –

    The full path to the root directory of your application.

    The project_root is used to mark frames in a stack trace either as being in your application or outside of the application.

  • in_app_include (List[str]) –

    A list of string prefixes of module names that belong to the app.

    This option takes precedence over in_app_exclude.

    Sentry differentiates stack frames that are directly related to your application (“in application”) from stack frames that come from other packages such as the standard library, frameworks, or other dependencies.

    The application package is automatically marked as inApp.

    The difference is visible in [sentry.io](https://sentry.io), where only the “in application” frames are displayed by default.

  • in_app_exclude (List[str]) –

    A list of string prefixes of module names that do not belong to the app, but rather to third-party packages.

    Modules considered not part of the app will be hidden from stack traces by default.

    This option can be overridden using in_app_include.

  • max_request_body_size (str) –

    This parameter controls whether integrations should capture HTTP request bodies. It can be set to one of the following values:

    • never: Request bodies are never sent.

    • small: Only small request bodies will be captured. The cutoff for small depends on the SDK (typically 4KB).

    • medium: Medium and small requests will be captured (typically 10KB).

    • always: The SDK will always capture the request body as long as Sentry can make sense of it.

    Please note that the Sentry server [limits HTTP request body size](https://develop.sentry.dev/sdk/ expected-features/data-handling/#variable-size). The server always enforces its size limit, regardless of how you configure this option.

  • max_value_length (int) –

    The number of characters after which the values containing text in the event payload will be truncated.

    WARNING: If the value you set for this is exceptionally large, the event may exceed 1 MiB and will be dropped by Sentry.

  • ca_certs (Optional[str]) – A path to an alternative CA bundle file in PEM-format.

  • send_client_reports (bool) –

    Set this boolean to False to disable sending of client reports.

    Client reports allow the client to send status reports about itself to Sentry, such as information about events that were dropped before being sent.

  • integrations (Sequence[Integration]) –

    List of integrations to enable in addition to auto-enabling integrations (overview).

    This setting can be used to override the default config options for a specific auto-enabling integration or to add an integration that is not auto-enabled.

  • disabled_integrations (Optional[Sequence[Integration]]) –

    List of integrations that will be disabled.

    This setting can be used to explicitly turn off specific auto-enabling integrations (list) or default integrations.

  • auto_enabling_integrations (bool) –

    Configures whether auto-enabling integrations (configuration) should be enabled.

    When set to False, no auto-enabling integrations will be enabled by default, even if the corresponding framework/library is detected.

  • default_integrations (bool) –

    Configures whether default integrations should be enabled.

    Setting default_integrations to False disables all default integrations as well as all auto-enabling integrations, unless they are specifically added in the integrations option, described above.

  • before_send (Optional[Callable[[Event, Dict[str, Any]], Optional[Event]]]) –

    This function is called with an SDK-specific message or error event object, and can return a modified event object, or null to skip reporting the event.

    This can be used, for instance, for manual PII stripping before sending.

    By the time before_send is executed, all scope data has already been applied to the event. Further modification of the scope won’t have any effect.

  • before_send_transaction (Optional[Callable[[Event, Dict[str, Any]], Optional[Event]]]) –

    This function is called with an SDK-specific transaction event object, and can return a modified transaction event object, or null to skip reporting the event.

    One way this might be used is for manual PII stripping before sending.

  • before_breadcrumb (Optional[Callable[[Dict[str, Any], Dict[str, Any]], Optional[Dict[str, Any]]]]) –

    This function is called with an SDK-specific breadcrumb object before the breadcrumb is added to the scope.

    When nothing is returned from the function, the breadcrumb is dropped.

    To pass the breadcrumb through, return the first argument, which contains the breadcrumb object.

    The callback typically gets a second argument (called a “hint”) which contains the original object from which the breadcrumb was created to further customize what the breadcrumb should look like.

  • transport (Union[Transport, Type[Transport], Callable[[Event], None], None]) –

    Switches out the transport used to send events.

    How this works depends on the SDK. It can, for instance, be used to capture events for unit-testing or to send it through some more complex setup that requires proxy authentication.

  • transport_queue_size (int) – The maximum number of events that will be queued before the transport is forced to flush.

  • http_proxy (Optional[str]) –

    When set, a proxy can be configured that should be used for outbound requests.

    This is also used for HTTPS requests unless a separate https_proxy is configured. However, not all SDKs support a separate HTTPS proxy.

    SDKs will attempt to default to the system-wide configured proxy, if possible. For instance, on Unix systems, the http_proxy environment variable will be picked up.

  • https_proxy (Optional[str]) –

    Configures a separate proxy for outgoing HTTPS requests.

    This value might not be supported by all SDKs. When not supported the http-proxy value is also used for HTTPS requests at all times.

  • proxy_headers (Optional[Dict[str, str]]) – A dict containing additional proxy headers (usually for authentication) to be forwarded to urllib3’s ProxyManager.

  • shutdown_timeout (float) –

    Controls how many seconds to wait before shutting down.

    Sentry SDKs send events from a background queue. This queue is given a certain amount to drain pending events. The default is SDK specific but typically around two seconds.

    Setting this value too low may cause problems for sending events from command line applications.

    Setting the value too high will cause the application to block for a long time for users experiencing network connectivity problems.

  • keep_alive (Optional[bool]) –

    Determines whether to keep the connection alive between requests.

    This can be useful in environments where you encounter frequent network issues such as connection resets.

  • cert_file (Optional[str]) –

    Path to the client certificate to use.

    If set, supersedes the CLIENT_CERT_FILE environment variable.

  • key_file (Optional[str]) –

    Path to the key file to use.

    If set, supersedes the CLIENT_KEY_FILE environment variable.

  • socket_options (Optional[List[Tuple[int, int, int | bytes]]]) –

    An optional list of socket options to use.

    These provide fine-grained, low-level control over the way the SDK connects to Sentry.

    If provided, the options will override the default urllib3 socket options.

  • traces_sample_rate (Optional[float]) –

    A number between 0 and 1, controlling the percentage chance a given transaction will be sent to Sentry.

    (0 represents 0% while 1 represents 100%.) Applies equally to all transactions created in the app.

    Either this or traces_sampler must be defined to enable tracing.

    If traces_sample_rate is 0, this means that no new traces will be created. However, if you have another service (for example a JS frontend) that makes requests to your service that include trace information, those traces will be continued and thus transactions will be sent to Sentry.

    If you want to disable all tracing you need to set traces_sample_rate=None. In this case, no new traces will be started and no incoming traces will be continued.

  • traces_sampler (Optional[Callable[[Dict[str, Any]], Union[float, int, bool]]]) –

    A function responsible for determining the percentage chance a given transaction will be sent to Sentry.

    It will automatically be passed information about the transaction and the context in which it’s being created, and must return a number between 0 (0% chance of being sent) and 1 (100% chance of being sent).

    Can also be used for filtering transactions, by returning 0 for those that are unwanted.

    Either this or traces_sample_rate must be defined to enable tracing.

  • trace_propagation_targets (Optional[Sequence[str]]) –

    An optional property that controls which downstream services receive tracing data, in the form of a sentry-trace and a baggage header attached to any outgoing HTTP requests.

    The option may contain a list of strings or regex against which the URLs of outgoing requests are matched.

    If one of the entries in the list matches the URL of an outgoing request, trace data will be attached to that request.

    String entries do not have to be full matches, meaning the URL of a request is matched when it _contains_ a string provided through the option.

    If trace_propagation_targets is not provided, trace data is attached to every outgoing request from the instrumented client.

  • functions_to_trace (Sequence[Dict[str, str]]) –

    An optional list of functions that should be set up for tracing.

    For each function in the list, a span will be created when the function is executed.

    Functions in the list are represented as strings containing the fully qualified name of the function.

    This is a convenient option, making it possible to have one central place for configuring what functions to trace, instead of having custom instrumentation scattered all over your code base.

    To learn more, see the Custom Instrumentation documentation.

  • enable_backpressure_handling (bool) –

    When enabled, a new monitor thread will be spawned to perform health checks on the SDK.

    If the system is unhealthy, the SDK will keep halving the traces_sample_rate set by you in 10 second intervals until recovery.

    This down sampling helps ensure that the system stays stable and reduces SDK overhead under high load.

    This option is enabled by default.

  • enable_db_query_source (bool) – When enabled, the source location will be added to database queries.

  • db_query_source_threshold_ms (int) –

    The threshold in milliseconds for adding the source location to database queries.

    The query location will be added to the query for queries slower than the specified threshold.

  • custom_repr (Optional[Callable[..., Optional[str]]]) –

    A custom repr function to run while serializing an object.

    Use this to control how your custom objects and classes are visible in Sentry.

    Return a string for that repr value to be used or None to continue serializing how Sentry would have done it anyway.

  • profiles_sample_rate (Optional[float]) –

    A number between 0 and 1, controlling the percentage chance a given sampled transaction will be profiled.

    (0 represents 0% while 1 represents 100%.) Applies equally to all transactions created in the app.

    This is relative to the tracing sample rate - e.g. 0.5 means 50% of sampled transactions will be profiled.

  • profiles_sampler (Optional[Callable[[Dict[str, Any]], Union[float, int, bool]]])

  • profiler_mode (Union[Literal['thread', 'gevent', 'unknown'], Literal['sleep'], None])

  • profile_lifecycle (Literal['manual', 'trace'])

  • profile_session_sample_rate (Optional[float])

  • enable_tracing (Optional[bool])

  • propagate_traces (bool)

  • auto_session_tracking (bool)

  • spotlight (Union[bool, str, None])

  • instrumenter (Optional[str])

  • _experiments (Experiments)

Capturing Data

sentry_sdk.api.capture_event(event, hint=None, scope=None, **scope_kwargs)[source]

Alias for sentry_sdk.Scope.capture_event()

Captures an event.

Merges given scope data and calls sentry_sdk.client._Client.capture_event().

Parameters:

  • event (Event) – A ready-made event that can be directly sent to Sentry.

  • hint (Optional[Dict[str, Any]]) – Contains metadata about the event that can be read from before_send, such as the original exception object or a HTTP request object.

  • scope (Optional[Any]) – An optional sentry_sdk.Scope to apply to events. The scope and scope_kwargs parameters are mutually exclusive.

  • scope_kwargs (Any) – Optional data to apply to event. For supported **scope_kwargs see sentry_sdk.Scope.update_from_kwargs(). The scope and scope_kwargs parameters are mutually exclusive.

Return type:

Optional[str]

Returns:

An event_id if the SDK decided to send the event (see sentry_sdk.client._Client.capture_event()).

sentry_sdk.api.capture_exception(error=None, scope=None, **scope_kwargs)[source]

Alias for sentry_sdk.Scope.capture_exception()

Captures an exception.

Parameters:
Return type:

Optional[str]

Returns:

An event_id if the SDK decided to send the event (see sentry_sdk.client._Client.capture_event()).

sentry_sdk.api.capture_message(message, level=None, scope=None, **scope_kwargs)[source]

Alias for sentry_sdk.Scope.capture_message()

Captures a message.

Parameters:

  • message (str) – The string to send as the message.

  • level (Optional[Literal['fatal', 'critical', 'error', 'warning', 'info', 'debug']]) – If no level is provided, the default level is info.

  • scope (Optional[Any]) – An optional sentry_sdk.Scope to apply to events. The scope and scope_kwargs parameters are mutually exclusive.

  • scope_kwargs (Any) – Optional data to apply to event. For supported **scope_kwargs see sentry_sdk.Scope.update_from_kwargs(). The scope and scope_kwargs parameters are mutually exclusive.

Return type:

Optional[str]

Returns:

An event_id if the SDK decided to send the event (see sentry_sdk.client._Client.capture_event()).

Enriching Events

sentry_sdk.api.add_attachment(bytes=None, filename=None, path=None, content_type=None, add_to_transactions=False)[source]

Alias for sentry_sdk.Scope.add_attachment()

Adds an attachment to future events sent from this scope.

The parameters are the same as for the sentry_sdk.attachments.Attachment constructor.

Return type:

None

sentry_sdk.api.add_breadcrumb(crumb=None, hint=None, **kwargs)[source]

Alias for sentry_sdk.Scope.add_breadcrumb()

Adds a breadcrumb.

Parameters:

  • crumb (Optional[Dict[str, Any]]) – Dictionary with the data as the sentry v7/v8 protocol expects.

  • hint (Optional[Dict[str, Any]]) – An optional value that can be used by before_breadcrumb to customize the breadcrumbs that are emitted.

Return type:

None

sentry_sdk.api.set_context(key, value)[source]

Alias for sentry_sdk.Scope.set_context()

Binds a context at a certain key to a specific value.

Return type:

None

sentry_sdk.api.set_extra(key, value)[source]

Alias for sentry_sdk.Scope.set_extra()

Sets an extra key to a specific value.

Return type:

None

sentry_sdk.api.set_level(value)[source]

Alias for sentry_sdk.Scope.set_level()

Sets the level for the scope.

Parameters:

value (Literal['fatal', 'critical', 'error', 'warning', 'info', 'debug']) – The level to set.

Return type:

None

sentry_sdk.api.set_tag(key, value)[source]

Alias for sentry_sdk.Scope.set_tag()

Sets a tag for a key to a specific value.

Parameters:

  • key (str) – Key of the tag to set.

  • value (Any) – Value of the tag to set.

Return type:

None

sentry_sdk.api.set_user(value)[source]

Alias for sentry_sdk.Scope.set_user()

Sets a user for the scope.

Return type:

None

Performance Monitoring

sentry_sdk.api.continue_trace(environ_or_headers, op=None, name=None, source=None, origin='manual')[source]

Sets the propagation context from environment or headers and returns a transaction.

Return type:

Transaction

sentry_sdk.api.get_current_span(scope=None)[source]

Returns the currently active span if there is one running, otherwise None

Return type:

Optional[Span]

sentry_sdk.api.start_span(**kwargs)[source]

Alias for sentry_sdk.Scope.start_span()

Start a span whose parent is the currently active span or transaction, if any.

The return value is a sentry_sdk.tracing.Span instance, typically used as a context manager to start and stop timing in a with block.

Only spans contained in a transaction are sent to Sentry. Most integrations start a transaction at the appropriate time, for example for every incoming HTTP request. Use sentry_sdk.start_transaction() to start a new transaction when one is not already in progress.

For supported **kwargs see sentry_sdk.tracing.Span.

The instrumenter parameter is deprecated for user code, and it will be removed in the next major version. Going forward, it should only be used by the SDK itself.

Return type:

Span

sentry_sdk.api.start_transaction(transaction=None, instrumenter='sentry', custom_sampling_context=None, **kwargs)[source]

Alias for sentry_sdk.Scope.start_transaction()

Start and return a transaction.

Start an existing transaction if given, otherwise create and start a new transaction with kwargs.

This is the entry point to manual tracing instrumentation.

A tree structure can be built by adding child spans to the transaction, and child spans to other spans. To start a new child span within the transaction or any span, call the respective .start_child() method.

Every child span must be finished before the transaction is finished, otherwise the unfinished spans are discarded.

When used as context managers, spans and transactions are automatically finished at the end of the with block. If not using context managers, call the .finish() method.

When the transaction is finished, it will be sent to Sentry with all its finished child spans.

Parameters:

  • transaction (Optional[Transaction]) – The transaction to start. If omitted, we create and start a new transaction.

  • instrumenter (str) – This parameter is meant for internal use only. It will be removed in the next major version.

  • custom_sampling_context (Optional[Dict[str, Any]]) – The transaction’s custom sampling context.

  • kwargs (Unpack[TransactionKwargs]) – Optional keyword arguments to be passed to the Transaction constructor. See sentry_sdk.tracing.Transaction for available arguments.

Return type:

Union[Transaction, NoOpSpan]

Distributed Tracing

sentry_sdk.api.get_baggage()[source]

Returns Baggage either from the active span or from the scope.

Return type:

Optional[str]

sentry_sdk.api.get_traceparent()[source]

Returns the traceparent either from the active span or from the scope.

Return type:

Optional[str]

Client Management

sentry_sdk.api.is_initialized()[source]
Return type:

bool

Added in version 2.0.0.

Returns whether Sentry has been initialized or not.

If a client is available and the client is active (meaning it is configured to send data) then Sentry is initialized.

sentry_sdk.api.get_client()[source]

Alias for sentry_sdk.Scope.get_client() :rtype: BaseClient

Added in version 2.0.0.

Returns the currently used sentry_sdk.Client. This checks the current scope, the isolation scope and the global scope for a client. If no client is available a sentry_sdk.client.NonRecordingClient is returned.

Managing Scope (advanced)

sentry_sdk.api.configure_scope(callback=None)[source]

Reconfigures the scope.

Parameters:

callback (Optional[Callable[[Scope], None]]) – If provided, call the callback with the current scope.

Return type:

Optional[ContextManager[Scope]]

Returns:

If no callback is provided, returns a context manager that returns the scope.

sentry_sdk.api.push_scope(callback=None)[source]

Pushes a new layer on the scope stack.

Parameters:

callback (Optional[Callable[[Scope], None]]) – If provided, this method pushes a scope, calls callback, and pops the scope again.

Return type:

Optional[ContextManager[Scope]]

Returns:

If no callback is provided, a context manager that should be used to pop the scope again.

sentry_sdk.api.new_scope()[source]
Return type:

Generator[Scope, None, None]

Added in version 2.0.0.

Context manager that forks the current scope and runs the wrapped code in it. After the wrapped code is executed, the original scope is restored.

Example Usage:

import sentry_sdk

with sentry_sdk.new_scope() as scope:
    scope.set_tag("color", "green")
    sentry_sdk.capture_message("hello") # will include `color` tag.

sentry_sdk.capture_message("hello, again") # will NOT include `color` tag.