Expose the LLM response text from the Assist Pipeline as:

[ondrom111]

Describe the feature

There is currently no variable, event, or hook that contains the LLM response.
This means:

• The last spoken sentence cannot be saved to an entity
• It cannot be translated
• It cannot be displayed on a dashboard
• It cannot be logged
• It cannot be used in automations
• It cannot be processed further in any way

For example, if a user wants to say “translate” and have Home Assistant translate the last LLM reply, this is impossible because the reply text is not available anywhere.

Requested solution

Expose the LLM response text from the Assist Pipeline as:

• a variable (e.g., {{ conversation.response }}), or
• an event (e.g., conversation.response), or
• a way to run an action in the “LLM Response” step with access to the response variable.

Any of these would allow automations to work with the actual text returned by the LLM.

• Language learning (translate last sentence)
• Accessibility features
• Logging conversations
• Displaying responses on dashboards
• Sending responses via notifications
• Debugging LLM behavior
• Advanced automations based on LLM output

Thank you for considering this feature.

Example commands

For example, if a user wants to say “translate” and have Home Assistant translate the last LLM reply, this is impossible because the reply text is not available anywhere.

Use cases

Assist Pipeline already receives the LLM response internally, but Home Assistant does not expose it.
Making the response available to automations would unlock many powerful use cases and significantly improve the flexibility of the Assist system.

Anything else?

No response