Skip to main content

Data Entry Flow

Data Entry Flow is a data entry framework that is part of Home Assistant. Data entry is done via data entry flows. A flow can represent a simple login form or a multi-step setup wizard for a component. A Flow Manager manages all flows that are in progress and handles creation of new flows.

Data Entry Flow is used in Home Assistant to login, create config entries, handle options flow, repair issues.

Flow Manager

This is the class that manages the flows that are in progress. When instantiating one, you pass in two async callbacks:

async def async_create_flow(handler, context=context, data=data):
"""Create flow."""

The manager delegates instantiating of config flow handlers to this async callback. This allows the parent of the manager to define their own way of finding handlers and preparing a handler for instantiation. For example, in the case of the config entry manager, it will make sure that the dependencies and requirements are setup.

async def async_finish_flow(flow, result):
"""Finish flow."""

This async callback is called when a flow is finished or aborted. i.e. result['type'] in [FlowResultType.CREATE_ENTRY, FlowResultType.ABORT]. The callback function can modify result and return it back, if the result type changed to FlowResultType.FORM, the flow will continue running, display another form.

If the result type is FlowResultType.FORM, the result should look like:

{
# The result type of the flow
"type": FlowResultType.FORM,
# the id of the flow
"flow_id": "abcdfgh1234",
# handler name
"handler": "hue",
# name of the step, flow.async_step_[step_id] will be called when form submitted
"step_id": "init",
# a voluptuous schema to build and validate user input
"data_schema": vol.Schema(),
# an errors dict, None if no errors
"errors": errors,
# a detail information about the step
"description_placeholders": description_placeholders,
}

If the result type is FlowResultType.CREATE_ENTRY, the result should look like:

{
# Data schema version of the entry
"version": 2,
# The result type of the flow
"type": FlowResultType.CREATE_ENTRY,
# the id of the flow
"flow_id": "abcdfgh1234",
# handler name
"handler": "hue",
# title and data as created by the handler
"title": "Some title",
"result": {
"some": "data"
},
}

If the result type is FlowResultType.ABORT, the result should look like:

{
# The result type of the flow
"type": FlowResultType.ABORT,
# the id of the flow
"flow_id": "abcdfgh1234",
# handler name
"handler": "hue",
# the abort reason
"reason": "already_configured",
}

Flow Handler

Flow handlers will handle a single flow. A flow contains one or more steps. When a flow is instantiated, the FlowHandler.init_step step will be called. Each step has several possible results:

At a minimum, each flow handler will have to define a version number and a step. This doesn't have to be init, as async_create_flow can assign init_step dependent on the current workflow, for example in configuration, context.source will be used as init_step.

For example, a bare minimum config flow would be:

from homeassistant import data_entry_flow

@config_entries.HANDLERS.register(DOMAIN)
class ExampleConfigFlow(data_entry_flow.FlowHandler):

# The schema version of the entries that it creates
# Home Assistant will call your migrate method if the version changes
# (this is not implemented yet)
VERSION = 1

async def async_step_user(self, user_input=None):
"""Handle user step."""

Data entry flows depend on translations for showing the text in the steps. It depends on the parent of a data entry flow manager where this is stored. For config and option flows, this is in strings.json under config and option, respectively.

For a more detailed explanation of strings.json see the backend translation page.

Show Form

This result type will show a form to the user to fill in. You define the current step, the schema of the data (using a mixture of voluptuous and/or selectors) and optionally a dictionary of errors.

from homeassistant.helpers.selector import selector

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_user(self, user_input=None):
# Specify items in the order they are to be displayed in the UI
data_schema = {
vol.Required("username"): str,
vol.Required("password"): str,
}

if self.show_advanced_options:
data_schema["allow_groups"] = selector({
"select": {
"options": ["all", "light", "switch"],
}
})

return self.async_show_form(step_id="init", data_schema=vol.Schema(data_schema))

Labels & Descriptions

Translations for the form are added to strings.json in a key for the step_id. That object may contain the folowing keys:

KeyValueNotes
titleForm headingDo not include your brand name. It will be automatically injected from your manifest.
descriptionForm instructionsOptional. Do not link to the documentation as that is linked automatically. Do not include "basic" information like "Here you can set up X".
dataField labelsKeep succinct and consistent with other integrations whenever appropriate for the best user experience.
data_descriptionField descriptionsOptional explanatory text to show below the field.

The field labels and descriptions are given as a dictionary with keys corresponding to your schema. Here is a simple example:

{
"config": {
"step": {
"user": {
"title": "Add Group",
"description": "Some description",
"data": {
"entities": "Entities",
},
"data_description": {
"entities": "The entities to add to the group",
},
}
}
}
}

Enabling Browser Autofill

Suppose your integration is collecting form data which can be automatically filled by browsers or password managers, such as login credentials or contact information. You should enable autofill whenever possible for the best user experience and accessibility. There are two options to enable this.

The first option is to use Voluptuous with data keys recognized by the frontend. The frontend will recognize the keys "username" and "password" and add HTML autocomplete attribute values of "username" and "current-password" respectively. Support for autocomplete is limited to "username" and "password" fields and is supported primarily to quickly enable auto-fill on the many integrations that collect them without converting their schemas to selectors.

The second option is to use a text selector. A text selector gives full control of the input type and allows any permitted value for autocomplete to be specified. A hypothetical schema collecting specific fillable data might be:

import voluptuous as vol
from homeassistant.const import CONF_PASSWORD, CONF_USERNAME
from homeassistant.helpers.selector import (
TextSelector,
TextSelectorConfig,
TextSelectorType,
)

STEP_USER_DATA_SCHEMA = vol.Schema(
{
vol.Required(CONF_USERNAME): TextSelector(
TextSelectorConfig(type=TextSelectorType.EMAIL, autocomplete="username")
),
vol.Required(CONF_PASSWORD): TextSelector(
TextSelectorConfig(
type=TextSelectorType.PASSWORD, autocomplete="current-password"
)
),
vol.Required("postal_code"): TextSelector(
TextSelectorConfig(type=TextSelectorType.TEXT, autocomplete="postal-code")
),
vol.Required("mobile_number"): TextSelector(
TextSelectorConfig(type=TextSelectorType.TEL, autocomplete="tel")
),
}
)

Defaults & Suggestions

If you'd like to pre-fill data in the form, you have two options. The first is to use the default parameter. This will both pre-fill the field, and act as the default value in case the user leaves the field empty.

    data_schema = {
vol.Optional("field_name", default="default value"): str,
}

The other alternative is to use a suggested value - this will also pre-fill the form field, but will allow the user to leave it empty if the user so wishes.

    data_schema = {
vol.Optional(
"field_name", description={"suggested_value": "suggested value"}
): str,
}

You can also mix and match - pre-fill through suggested_value, and use a different value for default in case the field is left empty, but that could be confusing to the user so use carefully.

Using suggested values also make it possible to declare a static schema, and merge suggested values from existing input. A add_suggested_values_to_schema helper makes this possible:

OPTIONS_SCHEMA = vol.Schema(
{
vol.Optional("field_name", default="default value"): str,
}
)

class ExampleOptionsFlow(config_entries.OptionsFlow):
async def async_step_init(
self, user_input: dict[str, Any] | None = None
) -> FlowResult:
return self.async_show_form(
data_schema = self.add_suggested_values_to_schema(
OPTIONS_SCHEMA, self.entry.options
)
)

Validation

After the user has filled in the form, the step method will be called again and the user input is passed in. Your step will only be called if the user input passes your data schema. When the user passes in data, you will have to do extra validation of the data. For example, you can verify that the passed in username and password are valid.

If something is wrong, you can return a dictionary with errors. Each key in the error dictionary refers to a field name that contains the error. Use the key base if you want to show an error unrelated to a specific field. The specified errors need to refer to a key in a translation file.

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_user(self, user_input=None):
errors = {}
if user_input is not None:
# Validate user input
valid = await is_valid(user_input)
if valid:
# See next section on create entry usage
return self.async_create_entry(...)

errors["base"] = "auth_error"

# Specify items in the order they are to be displayed in the UI
data_schema = {
vol.Required("username"): str,
vol.Required("password"): str,
}

return self.async_show_form(
step_id="init", data_schema=vol.Schema(data_schema), errors=errors
)

Multi-step flows

If the user input passes validation, you can return one of the possible step types again. If you want to navigate the user to the next step, return the return value of that step:

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_init(self, user_input=None):
errors = {}
if user_input is not None:
# Validate user input
valid = await is_valid(user_input)
if valid:
# Store info to use in next step
self.init_info = user_input
# Return the form of the next step
return await self.async_step_account()

...

Create Entry

When the result is "Create Entry", an entry will be created and passed to the parent of the flow manager. A success message is shown to the user and the flow is finished. You create an entry by passing a title, data and optionally options. The title can be used in the UI to indicate to the user which entry it is. Data and options can be any data type, as long as they are JSON serializable. Options are used for mutable data, for example a radius. Whilst Data is used for immutable data that isn't going to change in an entry, for example location data.

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_user(self, user_input=None):
return self.async_create_entry(
title="Title of the entry",
data={
"username": user_input["username"],
"password": user_input["password"]
},
options={
"mobile_number": user_input["mobile_number"]
},
)

Note: A user can change their password, which technically makes it mutable data, but for changing authentication credentials, you use reauthentication, which can mutate the config entry data.

Abort

When a flow cannot be finished, you need to abort it. This will finish the flow and inform the user that the flow has finished. Reasons for a flow to not be able to finish can be that a device is already configured or not compatible with Home Assistant.

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_user(self, user_input=None):
return self.async_abort(reason="not_supported")

External Step & External Step Done

It is possible that a user needs to finish a config flow by doing actions on an external website. For example, setting up an integration by being redirected to an external webpage. This is commonly used by integrations that use OAuth2 to authorize a user.

The example is about config entries, but works with other parts that use data entry flows too.

The flow works as follows:

  1. The user starts config flow in Home Assistant.

  2. Config flow prompts the user to finish the flow on an external website.

  3. The user opens the external website.

  4. Upon completion of the external step, the user's browser will be redirected to a Home Assistant endpoint to deliver the response.

  5. The endpoint validates the response, and upon validation, marks the external step as done and returns JavaScript code to close the window: <script>window.close()</script>.

    To be able to route the result of the external step to the Home Assistant endpoint, you will need to make sure the config flow ID is included. If your external step is an OAuth2 flow, you can leverage the oauth2 state for this. This is a variable that is not interpreted by the authorization page but is passed as-is to the Home Assistant endpoint.

  6. The window closes and the Home Assistant user interface with the config flow will be visible to the user again.

  7. The config flow has automatically advanced to the next step when the external step was marked as done. The user is prompted with the next step.

Example configuration flow that includes an external step.

from homeassistant import config_entries

@config_entries.HANDLERS.register(DOMAIN)
class ExampleConfigFlow(data_entry_flow.FlowHandler):
VERSION = 1
data = None

async def async_step_user(self, user_input=None):
if not user_input:
return self.async_external_step(
step_id="user",
url=f"https://example.com/?config_flow_id={self.flow_id}",
)

self.data = user_input
return self.async_external_step_done(next_step_id="finish")

async def async_step_finish(self, user_input=None):
return self.async_create_entry(title=self.data["title"], data=self.data)

Avoid doing work based on the external step data before you return an async_mark_external_step_done. Instead, do the work in the step that you refer to as next_step_id when marking the external step done. This will give the user a better user experience by showing a spinner in the UI while the work is done.

If you do the work inside the authorize callback, the user will stare at a blank screen until that all of a sudden closes because the data has forwarded. If you do the work before marking the external step as done, the user will still see the form with the "Open external website" button while the background work is being done. That too is undesirable.

Example code to mark an external step as done:

from homeassistant import data_entry_flow


async def handle_result(hass, flow_id, data):
result = await hass.config_entries.async_configure(flow_id, data)

if result["type"] == data_entry_flow.FlowResultType.EXTERNAL_STEP_DONE:
return "success!"
else:
return "Invalid config flow specified"

Show Progress & Show Progress Done

If a data entry flow step needs a considerable amount of time to finish, we should inform the user about this.

The example is about config entries, but works with other parts that use data entry flows too.

The flow works as follows:

  1. The user starts the config flow in Home Assistant.
  2. The config flow creates an asyncio.Task to execute the long running task.
  3. The config flow informs the user that a task is in progress and will take some time to finish by calling async_show_progress, passing the asyncio.Task object to it. The flow should pass a task specific string as progress_action parameter to represent the translated text string for the prompt.
  4. The config flow will be automatically called once the task is finished, but may also be called before the task has finished, for example if frontend reloads.
  • If the task is not yet finished, the flow should not create another task, but instead call async_show_progress again.
  • If the task is finished, the flow must call the async_show_progress_done, indicating the next step
  1. The frontend will update each time we call show progress or show progress done.
  2. The config flow will automatically advance to the next step when the progress was marked as done. The user is prompted with the next step.

Example configuration flow that includes two show sequential progress tasks.

import asyncio

from homeassistant import config_entries
from .const import DOMAIN

class TestFlow(config_entries.ConfigFlow, domain=DOMAIN):
VERSION = 1
task_one: asyncio.Task | None = None
task_two: asyncio.Task | None = None

async def async_step_user(self, user_input=None):
uncompleted_task: asyncio.Task[None] | None = None

if not self.task_one:
coro = asyncio.sleep(10)
self.task_one = self.hass.async_create_task(coro)
if not self.task_one.done():
progress_action = "task_one"
uncompleted_task = self.task_one
if not uncompleted_task:
if not self.task_two:
coro = asyncio.sleep(10)
self.task_two = self.hass.async_create_task(coro)
if not self.task_two.done():
progress_action = "task_two"
uncompleted_task = self.task_two
if uncompleted_task:
return self.async_show_progress(
progress_action=progress_action,
progress_task=uncompleted_task,
)

return self.async_show_progress_done(next_step_id="finish")

async def async_step_finish(self, user_input=None):
if not user_input:
return self.async_show_form(step_id="finish")
return self.async_create_entry(title="Some title", data={})

Show Menu

This will show a navigation menu to the user to easily pick the next step. The menu labels can be hardcoded by specifying a dictionary of {step_id: label} or translated via strings.json when specifying a list.

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_user(self, user_input=None):
return self.async_show_menu(
step_id="user",
menu_options=["discovery", "manual"],
description_placeholders={
"model": "Example model",
}
)
# Example showing the other approach
return self.async_show_menu(
step_id="user",
menu_options={
"option_1": "Option 1",
"option_2": "Option 2",
}
)
{
"config": {
"step": {
"user": {
"menu_options": {
"discovery": "Discovery",
"manual": "Manual ({model})",
}
}
}
}
}

Initializing a config flow from an external source

You might want to initialize a config flow programmatically. For example, if we discover a device on the network that requires user interaction to finish setup. To do so, pass a source parameter and optional user input when initializing a flow:

await flow_mgr.async_init(
"hue", context={"source": data_entry_flow.SOURCE_DISCOVERY}, data=discovery_info
)

The config flow handler will not start with the init step. Instead, it will be instantiated with a step name equal to the source. The step should follow the same return values as a normal step.

class ExampleConfigFlow(data_entry_flow.FlowHandler):
async def async_step_discovery(self, info):
"""Handle discovery info."""

The source of a config flow is available as self.source on FlowHandler.