Generate data
The Generate data action uses AI to run a task that generates data, and returns the result as response data. You give it a set of instructions, and the AI task entity generates either free text or structured data that you can use elsewhere, for example in a template sensor or a notification.
This action does not support targets. Instead, you describe the task and, optionally, choose which AI task entity to run it on. When you leave the entity out, the preferred AI task entity is used.
Using this action from the user interface
If you prefer building automations and scripts visually, Home Assistant walks you through this action step by step. You pick what to target, tweak a few options, and save. No YAML knowledge required.
To generate data from an automation or a script:
- Go to Settings > Automations & scenes.
- Open an existing automation or script, or select Create automation > Create new automation.
- If you’re setting up a new automation, add a trigger in the When section. Scripts don’t need a trigger. They run when something else calls them.
- In the Then do section, select Add action.
- From the search box, search for and select AI Task: Generate data.
- Enter a Task name and the Instructions, and set any of the options you need.
- In the Response variable field, enter a name to store the result, for example,
generated_text. - Select Save.
Options in the UI
A name that identifies the type of task, such as “home summary” or “alert notification”.
The AI task entity to run the task on. When left empty, the preferred AI task entity is used.
When set, the AI returns structured data with the fields you define. Each field can have a description, a selector, and an optional required property.
Using this action in YAML
If you work directly in YAML, or you want to know exactly what Home Assistant does under the hood, this section has the technical reference. It lists the field names you use in YAML, their types, and which ones are required.
In YAML, refer to this action as ai_task.generate_data. A basic example looks like this:
action: ai_task.generate_data
data:
task_name: "garage door left open comment"
instructions: "Generate a funny notification that the garage door was left open"
response_variable: generated_text
This generates the text and stores it in the generated_text response variable.
Options in YAML
A name that identifies the type of task, such as “home summary” or “alert notification”.
The specific instructions for the AI to follow when generating the data.
The AI task entity to run the task on. When left empty, the preferred AI task entity is used.
When set, the AI returns structured data with the fields you define. The structure is a mapping where the keys are the field names and the values contain a description, a selector, and an optional required property.
A list of files to attach for multi-modal AI analysis. Each attachment is the output of the Media selector.
Response data
The response data is a mapping with the following fields:
-
data: The generated data. This is free text, or structured data when you set thestructureoption. -
conversation_id: The ID of the conversation used for the task.
Try it yourself
Ready to test this? Open Developer tools > Actions, search for this action, fill in the fields, and select Perform action. You see what happens on your actual entitiesAn entity represents a sensor, actor, or function in Home Assistant. Entities are used to monitor physical properties or to control other entities. An entity is usually part of a device or a service. [Learn more] without writing a line of YAML.
Still stuck?
The Home Assistant community is quick to help: join Discord for real-time chat, post on the community forum with the action you’re calling and what you expected to happen, or share on our subreddit /r/homeassistant.
AI assistants like ChatGPT or Claude can also explain actions or suggest the right one when you describe what you want in plain language.
Related actions
These actions work well alongside this one:
- Generate image: Uses AI to generate an image from a set of instructions.