home-assistant.io/source/_docs/configuration/templating.markdown

title	description
Templating	Instructions on how to use the templating feature of Home Assistant.

This is an advanced feature of Home Assistant. You'll need a basic understanding of:

• Home Assistant architecture, especially states.
• The State object.

Templating is a powerful feature that allows you to control information going into and out of the system. It is used for:

• Formatting outgoing messages in, for example, the notify platforms and Alexa component.
• Process incoming data from sources that provide raw data, like MQTT, rest sensor or the command_line sensor.
• Automation Templating.

Building templates

Templating in Home Assistant is powered by the Jinja2 templating engine. This means that we are using their syntax and make some custom Home Assistant variables available to templates during rendering. Jinja2 supports a wide variety of operations:

• Mathematical operation
• Comparisons
• Logic

We will not go over the basics of the syntax, as Jinja2 does a great job of this in their templates documentation.

The frontend has a {% my developer_template title="template editor tool" %} to help develop and debug templates. Navigate to {% my developer_template title="Developer Tools > Template" %}, create your template in the Template editor and check the results on the right.

Templates can get big pretty fast. To keep a clear overview, consider using YAML multiline strings to define your templates:

{% raw %}

script:
  msg_who_is_home:
    sequence:
      - service: notify.notify
        data:
          message: >
            {% if is_state('device_tracker.paulus', 'home') %}
              Ha, Paulus is home!
            {% else %}
              Paulus is at {{ states('device_tracker.paulus') }}.
            {% endif %}

{% endraw %}

Important Template Rules

There are a few very important rules to remember when adding templates to YAML:

1. You must surround single-line templates with double quotes (") or single quotes (').
2. It is advised that you prepare for undefined variables by using if ... is not none or the default filter, or both.
3. It is advised that when comparing numbers, you convert the number(s) to a float or an int by using the respective filter.
4. While the float and int filters do allow a default fallback value if the conversion is unsuccessful, they do not provide the ability to catch undefined variables.

Remembering these simple rules will help save you from many headaches and endless hours of frustration when using automation templates.

Home Assistant template extensions

Extensions allow templates to access all of the Home Assistant specific states and adds other convenience functions and filters.

Limited Templates

Templates for some triggers as well as trigger_variables only support a subset of the Home Assistant template extensions. This subset is referred to as "Limited Templates".

States

Not supported in limited templates.

• Iterating states will yield each state sorted alphabetically by entity ID.
• Iterating states.domain will yield each state of that domain sorted alphabetically by entity ID.
• states.sensor.temperature returns the state object for sensor.temperature (avoid when possible, see note below).
• states('device_tracker.paulus') will return the state string (not the object) of the given entity, unknown if it doesn't exist, unavailable if the object exists but is not yet available.
• is_state('device_tracker.paulus', 'home') will test if the given entity is the specified state.
• state_attr('device_tracker.paulus', 'battery') will return the value of the attribute or None if it doesn't exist.
• is_state_attr('device_tracker.paulus', 'battery', 40) will test if the given entity attribute is the specified state (in this case, a numeric value). Note that the attribute can be None and you want to check if it is None, you need to use state_attr('sensor.my_sensor', 'attr') == None.

Avoid using states.sensor.temperature.state, instead use states('sensor.temperature'). It is strongly advised to use the states(), is_state(), state_attr() and is_state_attr() as much as possible, to avoid errors and error message when the entity isn't ready yet (e.g., during Home Assistant startup).

Besides the normal state object methods and properties, states.sensor.temperature.state_with_unit will print the state of the entity and, if available, the unit.

States examples

The next two statements result in the same value if the state exists. The second one will result in an error if the state does not exist.

{% raw %}

{{ states('device_tracker.paulus') }}
{{ states.device_tracker.paulus.state }}

{% endraw %}

Print out a list of all the sensor states:

{% raw %}

{% for state in states.sensor %}
  {{ state.entity_id }}={{ state.state }},
{% endfor %}

{% endraw %}

Other state examples: {% raw %}

{% if is_state('device_tracker.paulus', 'home') %}
  Ha, Paulus is home!
{% else %}
  Paulus is at {{ states('device_tracker.paulus') }}.
{% endif %}

#check sensor.train_departure_time state
{% if states('sensor.train_departure_time') in ("unavailable", "unknown") %}
  {{ ... }}

{% set state = states('sensor.temperature') %}{{ state | float + 1 if is_number(state) else "invalid temperature" }}

{% set state = states('sensor.temperature') %}{{ (state | float * 10) | round(2) if is_number(state)}}

{% set state = states('sensor.temperature') %}
{% if is_number(state) and state | float > 20 %}
  It is warm!
{% endif %}

{{ as_timestamp(states.binary_sensor.garage_door.last_changed) }}

{{ as_local(states.binary_sensor.garage_door.last_changed) }}

{{ as_timestamp(now()) - as_timestamp(states.binary_sensor.garage_door.last_changed) }}

{{ as_local(states.sensor.time.last_changed) }}

{{ states('sensor.expires') | as_datetime }}

{% endraw %}

Attributes

Not supported in limited templates.

You can print an attribute with state_attr if state is defined.

Attributes examples

{% raw %}

{% if states.device_tracker.paulus %}
  {{ state_attr('device_tracker.paulus', 'battery') }}
{% else %}
  ??
{% endif %}

{% endraw %}

With strings:

{% raw %}

{% set tracker_name = "paulus"%}

{% if states("device_tracker." + tracker_name) != "unknown" %}
  {{ state_attr("device_tracker." + tracker_name, "battery")}}
{% else %}
  ??
{% endif %}

{% endraw %}

Working with Groups

Not supported in limited templates.

The expand function and filter can be used to sort entities and expand groups. It outputs a sorted array of entities with no duplicates.

Expand examples

{% raw %}

{% for tracker in expand('device_tracker.paulus', 'group.child_trackers') %}
  {{ state_attr(tracker.entity_id, 'battery') }}
  {%- if not loop.last %}, {% endif -%}
{% endfor %}

{% endraw %}

The same thing can also be expressed as a filter:

{% raw %}

{{ expand(['device_tracker.paulus', 'group.child_trackers']) 
  | selectattr("attributes.battery", 'defined')
  | join(', ', attribute="attributes.battery") }}

{% endraw %}

{% raw %}

{% for energy in expand('group.energy_sensors') if is_number(energy.state) %}
  {{ energy.state }}
  {%- if not loop.last %}, {% endif -%}
{% endfor %}

{% endraw %}

The same thing can also be expressed as a test:

{% raw %}

{{ expand('group.energy_sensors') 
  | selectattr("state", 'is_number') | join(', ') }}

{% endraw %}

Devices

• device_entities(device_id) returns a list of entities that are associated with a given device ID. Can also be used as a filter.
• device_attr(device_or_entity_id, attr_name) returns the value of attr_name for the given device or entity ID. Not supported in limited templates.
• is_device_attr(device_or_entity_id, attr_name, attr_value) returns whether the value of attr_name for the given device or entity ID matches attr_value. Not supported in limited templates.
• device_id(entity_id) returns the device ID for a given entity ID or device name. Can also be used as a filter.

Devices examples

{% raw %}

{{ device_attr('deadbeefdeadbeefdeadbeefdeadbeef', 'manufacturer') }}  # Sony

{{ is_device_attr('deadbeefdeadbeefdeadbeefdeadbeef', 'manufacturer', 'Sony') }}  # true

{{ device_id('sensor.sony') }}  # deadbeefdeadbeefdeadbeefdeadbeef

{% endraw %}

Config Entries

• entry_id(entity_id) returns the config entry ID for a given entity ID. Can also be used as a filter.

Config entries examples

{% raw %}

{{ entry_id('sensor.sony') }}  # deadbeefdeadbeefdeadbeefdeadbeef

{% endraw %}

Areas

• area_id(lookup_value) returns the area ID for a given device ID, entity ID, or area name. Can also be used as a filter.
• area_name(lookup_value) returns the area name for a given device ID, entity ID, or area ID. Can also be used as a filter.
• area_entities(area_name_or_id) returns the list of entity IDs tied to a given area ID or name. Can also be used as a filter.
• area_devices(area_name_or_id) returns the list of device IDs tied to a given area ID or name. Can also be used as a filter.

Areas examples

{% raw %}

{{ area_id('Living Room') }}  # 'deadbeefdeadbeefdeadbeefdeadbeef'

{{ area_id('my_device_id') }}  # 'deadbeefdeadbeefdeadbeefdeadbeef'

{{ area_id('sensor.sony') }}  # 'deadbeefdeadbeefdeadbeefdeadbeef'

{{ area_name('deadbeefdeadbeefdeadbeefdeadbeef') }}  # 'Living Room'

{{ area_name('my_device_id') }}  # 'Living Room'

{{ area_name('sensor.sony') }}  # 'Living Room'

{{ area_entities('deadbeefdeadbeefdeadbeefdeadbeef') }}  # ['sensor.sony']

{{ area_devices('Living Room') }}  # ['my_device_id']

{% endraw %}

Integrations

• integration_entities(integration) returns a list of entities that are associated with a given integration, such as hue or zwave_js.
• integration_entities(title) if you have multiple instances set-up for an integration, you can also use the title you've set for the integration in case you only want to target a specific device bridge.

Integrations examples

{% raw %}

{{ integration_entities('hue') }}  # ['light.hue_light_upstairs', 'light.hue_light_downstairs']

{{ integration_entities('Hue bridge downstairs') }}  # ['light.hue_light_downstairs']

{% endraw %}

Immediate if (iif)

A common case is to conditionally return a value based on another value. For example, return a "Yes" or "No" when the light is on or off.

This can be written as:

{% raw %}

{% if is_state('light.kitchen', 'on') %}
  Yes
{% else %}
  No
{% endif %}

{% endraw %}

Or using a shorter syntax:

{% raw %}

{{ 'Yes' if is_state('light.kitchen', 'on') else 'No' }}

{% endraw %}

Additionally, to the above, you can use the iif function/filter, which is an immediate if.

Syntax: iif(condition, if_true, if_false, if_none)

iif returns the value of if_true if the condition is truthy, the value of if_false if it's falsy and the value of if_none if it's None. An empty string, an empty mapping or an an empty list, are all falsy, refer to the Python documentation for an in depth explanation.

if_true is optional, if it's omitted True is returned if the condition is truthy. if_false is optional, if it's omitted False is returned if the condition is falsy. if_none is optional, if it's omitted the value of if_false is returned if the condition is None.

Examples using iif:

{% raw %}

{{ iif(is_state('light.kitchen', 'on'), 'Yes', 'No') }}

{{ is_state('light.kitchen', 'on') | iif('Yes', 'No') }}

{{ (states('light.kitchen') == 'on') | iif('Yes', 'No') }}

The immediate if filter does not short-circuit like you might expect with a typical conditional statement. The if_true, if_false and if_none expressions will all be evaluated and the filter will simply return one of the resulting values. This means you cannot use this filter to prevent executing an expression which would result in an error.

For example, if you wanted to select a field from trigger in an automation based on the platform you might go to make this template: trigger.platform == 'event' | iif(trigger.event.data.message, trigger.to_state.state). This won't work because both expressions will be evaluated and one will fail since the field doesn't exist. Instead you have to do this trigger.event.data.message if trigger.platform == 'event' else trigger.to_state.state. This form of the expression short-circuits so if the platform is event the expression trigger.to_state.state will never be evaluated and won't cause an error.

{% endraw %}

Time

now() and utcnow() are not supported in limited templates.

• now() returns a datetime object that represents the current time in your time zone.

  • You can also use: now().second, now().minute, now().hour, now().day, now().month, now().year, now().weekday() and now().isoweekday() and other datetime attributes and functions.
  • Using now() will cause templates to be refreshed at the start of every new minute.

• utcnow() returns a datetime object of the current time in the UTC timezone.

  • For specific values: utcnow().second, utcnow().minute, utcnow().hour, utcnow().day, utcnow().month, utcnow().year, utcnow().weekday() and utcnow().isoweekday().
  • Using utcnow() will cause templates to be refreshed at the start of every new minute.

• today_at(value) converts a string containing a military time format to a datetime object with today's date in your time zone.

  {% raw %}

  # Is the current time past 10:15?
  {{ now() > today_at("10:15") }} 
  

  {% endraw %}

• as_datetime() converts a string containing a timestamp, or valid UNIX timestamp, to a datetime object.

• as_timestamp(value, default) converts datetime object or string to UNIX timestamp. If that fails, returns the default value, or if omitted raises an error. This function can also be used as a filter.

• as_local() converts datetime object to local time. This function can also be used as a filter.

• strptime(string, format) parses a string based on a format and returns a datetime object. If that fails, returns the default value, or if omitted raises an error.

• relative_time converts datetime object to its human-friendly "age" string. The age can be in second, minute, hour, day, month or year (but only the biggest unit is considered, e.g., if it's 2 days and 3 hours, "2 days" will be returned). Note that it only works for dates in the past.

• timedelta returns a timedelta object and accepts the same arguments as the Python datetime.timedelta function -- days, seconds, microseconds, milliseconds, minutes, hours, weeks.

  {% raw %}

  # 77 minutes before current time. 
  {{ now() - timedelta( hours = 1, minutes = 17 ) }} 
  

  {% endraw %}

• as_timedelta(string) converts a string to a timedelta object. Expects data in the format DD HH:MM:SS.uuuuuu, DD HH:MM:SS,uuuuuu, or as specified by ISO 8601 (e.g. P4DT1H15M20S which is equivalent to 4 1:15:20) or PostgreSQL’s day-time interval format (e.g. 3 days 04:05:06) This function can also be used as a filter.

  {% raw %}

  # Renders to "00:10:00" 
  {{ as_timedelta("PT10M") }} 
  

  {% endraw %}

• Filter timestamp_local(default) converts a UNIX timestamp to the ISO format string representation as date/time in your local timezone. If that fails, returns the default value, or if omitted raises an error. If a custom string format is needed in the string, use timestamp_custom instead.

• Filter timestamp_utc(default) converts a UNIX timestamp to the ISO format string representation representation as date/time in UTC timezone. If that fails, returns the default value, or if omitted raises an error. If a custom string format is needed in the string, use timestamp_custom instead.

• Filter timestamp_custom(format_string, local=True, default) converts an UNIX timestamp to its string representation based on a custom format, the use of a local timezone is the default. If that fails, returns the default value, or if omitted raises an error. Supports the standard Python time formatting options.

UNIX timestamp is the number of seconds that have elapsed since 00:00:00 UTC on 1 January 1970. Therefore, if used as a function's argument, it can be substituted with a numeric value (int or float).

If your template is returning a timestamp that should be displayed in the frontend (e.g., as a sensor entity with device_class: timestamp), you have to ensure that it is the ISO 8601 format (meaning it has the "T" separator between the date and time portion). Otherwise, frontend rendering on macOS and iOS devices will show an error. The following value template would result in such an error:

{% raw %}

{{ states.sun.sun.last_changed }} => 2021-01-24 07:06:59+00:00 (missing "T" separator)

{% endraw %}

To fix it, enforce the ISO conversion via isoformat():

{% raw %}

{{ states.sun.sun.last_changed.isoformat() }} => 2021-01-24T07:06:59+00:00 (contains "T" separator)

{% endraw %}

{% raw %}

{{ 120 | timestamp_local }}

{% endraw %}

To/From JSON

The to_json filter serializes an object to a JSON string. In some cases, it may be necessary to format a JSON string for use with a webhook, as a parameter for command-line utilities or any number of other applications. This can be complicated in a template, especially when dealing with escaping special characters. Using the to_json filter, this is handled automatically.

Similarly to the Python equivalent, the filter accepts an ensure_ascii parameter, defaulting to True. If ensure_ascii is True, the output is guaranteed to have all incoming non-ASCII characters escaped. If ensure_ascii is false, these characters will be output as-is.

The from_json filter operates similarly, but in the other direction, de-serializing a JSON string back into an object.

To/From JSON examples

In this example, the special character '°' will be automatically escaped in order to produce valid JSON. The difference between the stringified object and the actual JSON is evident.

Template

{% raw %}

{% set temp = {'temperature': 25, 'unit': '°C'} %}
stringified object: {{ temp }}
object|to_json: {{ temp|to_json(ensure_ascii=False) }}

{% endraw %}

Output

{% raw %}

stringified object: {'temperature': 25, 'unit': '°C'}
object|to_json: {"temperature": 25, "unit": "\u00b0C"}

{% endraw %}

Conversely, from_json can be used to de-serialize a JSON string back into an object to make it possible to easily extract usable data.

Template

{% raw %}

{% set temp = '{"temperature": 25, "unit": "\u00b0C"}'|from_json %}
The temperature is {{ temp.temperature }}{{ temp.unit }}

{% endraw %}

Output

{% raw %}

The temperature is 25°C

{% endraw %}

Is defined

Sometimes a template should only return if a value or object is defined, if not, the supplied default value should be returned. This can be useful to validate a JSON payload. The is_defined filter allows to throw an error if a value or object is not defined.

Example using is_defined to parse a JSON payload:

{% raw %}

{{ value_json.val | is_defined }}

{% endraw %}

This will throw an error UndefinedError: 'value_json' is undefined if the JSON payload has no val attribute.

Version

• version() Returns a AwesomeVersion object for the value given inside the brackets.
  • This is also available as a filter (| version).

Examples:

{% raw %}

• {{ version("2099.9.9") > "2000.0.0" }} Will return True
• {{ version("2099.9.9") < "2099.10" }} Will return True
• {{ "2099.9.9" | version < "2099.10" }} Will return True
• {{ (version("2099.9.9") - "2100.9.10").major }} Will return True
• {{ (version("2099.9.9") - "2099.10.9").minor }} Will return True
• {{ (version("2099.9.9") - "2099.9.10").patch }} Will return True

{% endraw %}

Distance

Not supported in limited templates.

• distance() will measure the distance in kilometers between home, entity, coordinates.
• closest() will find the closest entity.

Distance examples

If only one location is passed in, Home Assistant will measure the distance from home.

{% raw %}

Using Lat Lng coordinates: {{ distance(123.45, 123.45) }}

Using State: {{ distance(states.device_tracker.paulus) }}

These can also be combined in any combination:
{{ distance(123.45, 123.45, 'device_tracker.paulus') }}
{{ distance('device_tracker.anne_therese', 'device_tracker.paulus') }}

{% endraw %}

Closest examples

The closest function and filter will find the closest entity to the Home Assistant location:

{% raw %}

Query all entities: {{ closest(states) }}
Query all entities of a specific domain: {{ closest(states.device_tracker) }}
Query all entities in group.children: {{ closest('group.children') }}
Query all entities in group.children: {{ closest(states.group.children) }}

{% endraw %}

Find entities closest to a coordinate or another entity. All previous arguments still apply for second argument.

{% raw %}

Closest to a coordinate: {{ closest(23.456, 23.456, 'group.children') }}
Closest to an entity: {{ closest('zone.school', 'group.children') }}
Closest to an entity: {{ closest(states.zone.school, 'group.children') }}

{% endraw %}

Since closest returns a state, we can combine it with distance too.

{% raw %}

{{ closest(states).name }} is {{ distance(closest(states)) }} kilometers away.

{% endraw %}

The last argument of the closest function has an implicit expand, and can take any iterable sequence of states or entity IDs, and will expand groups:

{% raw %}

Closest out of given entities: 
    {{ closest(['group.children', states.device_tracker]) }}
Closest to a coordinate:  
    {{ closest(23.456, 23.456, ['group.children', states.device_tracker]) }}
Closest to some entity: 
    {{ closest(states.zone.school, ['group.children', states.device_tracker]) }}

{% endraw %}

It will also work as a filter over an iterable group of entities or groups:

{% raw %}

Closest out of given entities: 
    {{ ['group.children', states.device_tracker] | closest }}
Closest to a coordinate:  
    {{ ['group.children', states.device_tracker] | closest(23.456, 23.456) }}
Closest to some entity: 
    {{ ['group.children', states.device_tracker] | closest(states.zone.school) }}

{% endraw %}

Numeric functions and filters

Some of these functions can also be used in a filter. This means they can act as a normal function like this sqrt(2), or as part of a filter like this 2|sqrt.

The numeric functions and filters raise an error if the input is not a valid number, optionally a default value can be specified which will be returned instead. The is_number function and filter can be used to check if a value is a valid number. Errors can be caught by the default filter.

{% raw %}

• {{ float("not_a_number") }} - the template will fail to render
• {{ "not_a_number" | sin }} - the template will fail to render
• {{ float("not_a_number", default="Invalid number!") }} - renders as "Invalid number!"
• {{ "not_a_number" | sin(default="Invalid number!") }} - renders as "Invalid number!"

{% endraw %}

• float(value, default) function will attempt to convert the input to a float. If that fails, returns the default value, or if omitted raises an error.

• float(default) filter will attempt to convert the input to a float. If that fails, returns the default value, or if omitted raises an error.

• is_number will return True if the input can be parsed by Python's float function and the parsed input is not inf or nan, in all other cases returns False. Note that a Python bool will return True but the strings "True" and "False" will both return False. Can be used as a filter.

• int(value, default) function is similar to float, but converts to an int instead. Like float, it has a filter form, and an error is raised if the default value is omitted. Fractional part is discarded: int("1.5") is 1.

• bool(value, default) function converts the value to either true or false. The following values are considered to be true: boolean true, non-zero ints and floats, and the strings "true", "yes", "on", "enable", and "1" (case-insensitive). false is returned for the opposite values: boolean false, integer or floating-point 0, and the strings "false", "no", "off", "disable", and "0" (also case-insensitive). If the value is not listed here, the function returns the default value, or if omitted raises an error. This function is intended to be used on states of binary sensors, switches, or similar entities, so its behavior is different from Python's built-in bool conversion, which would consider e.g. "on", "off", and "unknown" all to be true, but "" to be false; if that is desired, use not not value or a similar construct instead. Like float and int, bool has a filter form. Using none as the default value is particularly useful in combination with the immediate if filter: it can handle all three possible cases in a single line.

• log(value, base, default) will take the logarithm of the input. When the base is omitted, it defaults to e - the natural logarithm. If value or base can't be converted to a float, returns the default value, or if omitted raises an error. Can also be used as a filter.

• sin(value, default) will return the sine of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• cos(value, default) will return the cosine of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• tan(value, default) will return the tangent of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• asin(value, default) will return the arcus sine of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• acos(value, default) will return the arcus cosine of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• atan(value, default) will return the arcus tangent of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• atan2(y, x, default) will return the four quadrant arcus tangent of y / x. If y or x can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• sqrt(value, default) will return the square root of the input. If value can't be converted to a float, returns the default value, or if omitted raises an error. Can be used as a filter.

• max([x, y, ...]) will obtain the largest item in a sequence. Uses the same parameters as the built-in max filter.

• min([x, y, ...]) will obtain the smallest item in a sequence. Uses the same parameters as the built-in min filter.

• average([x, y, ...], default) will return the average value of the sequence. If list is empty or contains non-numeric value, returns the default value, or if omitted raises an error. Can be used as a filter.

• e mathematical constant, approximately 2.71828.

• pi mathematical constant, approximately 3.14159.

• tau mathematical constant, approximately 6.28318.

• Filter round(precision, method, default) will convert the input to a number and round it to precision decimals. Round has four modes and the default mode (with no mode specified) will round-to-even. If the input value can't be converted to a float, returns the default value, or if omitted raises an error.

  • round(precision, "floor", default) will always round down to precision decimals
  • round(precision, "ceil", default) will always round up to precision decimals
  • round(1, "half", default) will always round to the nearest .5 value. precision should be 1 for this mode

• Filter value_one|bitwise_and(value_two) perform a bitwise and(&) operation with two values.

• Filter value_one|bitwise_or(value_two) perform a bitwise or(|) operation with two values.

• Filter ord will return for a string of length one an integer representing the Unicode code point of the character when the argument is a Unicode object, or the value of the byte when the argument is an 8-bit string.

Functions and filters to process raw data

These functions are used to process raw value's in a bytes format to values in a native Python type or vice-versa. The pack and unpack functions can also be used as a filter. They make use of the Python 3 struct library. See: https://docs.python.org/3/library/struct.html

• Filter value | pack(format_string) will convert a native type to a bytes type object. This will call function struct.pack(format_string, value). Returns None if an error occurs or when format_string is invalid.
• Function pack(value, format_string) will convert a native type to a bytes type object. This will call function struct.pack(format_string, value). Returns None if an error occurs or when format_string is invalid.
• Filter value | unpack(format_string, offset=0) will try to convert a bytes object into a native Python object. The offset parameter defines the offset position in bytes from the start of the input bytes based buffer. This will call function struct.unpack_from(format_string, value, offset=offset). Returns None if an error occurs or when format_string is invalid. Note that the filter unpack will only return the first bytes object, despite the function struct.unpack_from supporting to return multiple objects (e.g. with format_string being ">hh".
• Function unpack(value, format_string, offset=0) will try to convert a bytes object into a native Python object. The offset parameter defines the offset position in bytes from the start of the input bytes based buffer. This will call function struct.unpack_from(format_string, value, offset=offset). Returns None if an error occurs or when format_string is invalid. Note that the function unpack will only return the first bytes object, despite the function struct.unpack_from supporting to return multiple objects (e.g. with format_string being ">hh".

Some examples: {% raw %}

• {{ 0xDEADBEEF | pack(">I") }} - renders as b"\xde\xad\xbe\xef"
• {{ pack(0xDEADBEEF, ">I") }} - renders as b"\xde\xad\xbe\xef"
• {{ "0x%X" % 0xDEADBEEF | pack(">I") | unpack(">I") }} - renders as 0xDEADBEEF
• {{ "0x%X" % 0xDEADBEEF | pack(">I") | unpack(">H", offset=2) }} - renders as 0xBEEF

{% endraw %}

String filters

• Filter urlencode will convert an object to a percent-encoded ASCII text string (e.g., for HTTP requests using application/x-www-form-urlencoded).
• Filter slugify(separator="_") will convert a given string into a "slug".

Regular expressions

• Test string is match(find, ignorecase=False) will match the find expression at the beginning of the string using regex.
• Test string is search(find, ignorecase=False) will match the find expression anywhere in the string using regex.
• Filter string|regex_replace(find='', replace='', ignorecase=False) will replace the find expression with the replace string using regex.
• Filter value | regex_findall(find='', ignorecase=False) will find all regex matches of the find expression in value and return the array of matches.
• Filter value | regex_findall_index(find='', index=0, ignorecase=False) will do the same as regex_findall and return the match at index.

Processing incoming data

The other part of templating is processing incoming data. It allows you to modify incoming data and extract only the data you care about. This will only work for platforms and integrations that mention support for this in their documentation.

It depends per integration or platform, but it is common to be able to define a template using the value_template configuration key. When a new value arrives, your template will be rendered while having access to the following values on top of the usual Home Assistant extensions:

Variable	Description
value	The incoming value.
value_json	The incoming value parsed as JSON.

This means that if the incoming values looks like the sample below:

{
  "on": "true",
  "temp": 21
}

The template for on would be:

{% raw %}

'{{value_json.on}}'

{% endraw %}

Nested JSON in a response is supported as well:

{
  "sensor": {
    "type": "air",
    "id": "12345"
  },
  "values": {
    "temp": 26.09,
    "hum": 56.73
  }
}

Just use the "Square bracket notation" to get the value.

{% raw %}

"{{ value_json['values']['temp'] }}"

{% endraw %}

The following overview contains a couple of options to get the needed values:

{% raw %}

# Incoming value:
{"primes": [2, 3, 5, 7, 11, 13]}

# Extract first prime number 
{{ value_json.primes[0] }}

# Format output
{{ "%+.1f" | value_json }}

# Math
{{ value_json | float * 1024 if is_number(value_json) }}
{{ float(value_json) * (2**10) if is_number(value_json) }}
{{ value_json | log if is_number(value_json) }}
{{ log(1000, 10) }}
{{ sin(pi / 2) }}
{{ cos(tau) }}
{{ tan(pi) }}
{{ sqrt(e) }}

# Timestamps
{{ value_json.tst | timestamp_local }}
{{ value_json.tst | timestamp_utc }}
{{ value_json.tst | timestamp_custom('%Y', True) }}

{% endraw %}

To evaluate a response, go to {% my developer_template title="Developer Tools -> Template" %}, create your output in "Template editor", and check the result.

{% raw %}

{% set value_json=
    {"name":"Outside",
     "device":"weather-ha",
     "data":
        {"temp":"24C",
         "hum":"35%"
         } }%}

{{value_json.data.hum[:-1]}}

{% endraw %}

Using templates with the MQTT integration

The MQTT integration relies heavily on templates. Templates are used to transform incoming payloads (value templates) to status updates or incoming service calls (command templates) to payloads that configure the MQTT device.

Using value templates with MQTT

For incoming data a value template translates incoming JSON or raw data to a valid payload. Incoming payloads are rendered with possible JSON values, so when rendering the value_json can be used access the attributes in a JSON based payload.

Example value template:

With given payload:

{ "state": "ON", "temperature": 21.902 }

Template {% raw %}{{ value_json.temperature | round(1) }}{% endraw %} renders to 21.9.

Additional the MQTT entity attributes entity_id, name and this can be used as variables in the template. The this attribute refers to the entity state of the MQTT item.

Using command templates with MQTT

For service calls command templates are defined to format the outgoing MQTT payload to the device. When a service call is executed value can be used to generate the correct payload to the device.

Example command template:

With given value 21.9 template {% raw %}{"temperature": {{ value }} }{% endraw %} renders to:

{
  "temperature": 21.9
}

Additional the MQTT entity attributes entity_id, name and this can be used as variables in the template. The this attribute refers to the entity state of the MQTT item.

Some more things to keep in mind

entity_id that begins with a number

If your template uses an entity_id that begins with a number (example: states.device_tracker.2008_gmc) you must use a bracket syntax to avoid errors caused by rendering the entity_id improperly. In the example given, the correct syntax for the device tracker would be: states.device_tracker['2008_gmc']

Priority of operators

The default priority of operators is that the filter (|) has priority over everything except brackets. This means that:

{% raw %}

{{ states('sensor.temperature') | float / 10 | round(2) }}

{% endraw %}

Would round 10 to 2 decimal places, then divide states('sensor.temperature') by 10 (rounded to 2 decimal places so 10.00). This behavior is maybe not the one expected, but priority rules imply that.