Structured Output & New Timeline API

[valentinfrlch]

v1.6.0 Beta 1 Release Notes

This major release contains breaking changes, please read the breaking changes section before upgrading. It adds support for structured output and a new Timeline API.

⚠️ Breaking Changes

To avoid confusion with LLM Vision Memory:

• The remember action is now called create_event
• The remember parameter has been renamed to store_in_timeline.

Please make sure to update these in your custom automations. If you use the blueprint, make sure to update to the latest version.

If you use LLM Vision Cards, please update to v1.6.0 as well. The card now uses the new Timeline API, and older versions are no longer supported.

Contributors

A huge thank you to our contributors @rmaher001, @Smiley73, @muminkoykiran, @boian-ivanov, @ljmerza, @radzio, @TheRealFalseReality, and @dcgrove!

Integration

✨ Features

• Structured Output: Valid JSON output can now be forced using the structure parameter. See the docs for more details. (by @rmaher001)
• Timeline API: Events stored in the timeline can now be fetched through the new Timeline API. The LLM Vision Cards use this API, but you can also use it for your own automations or to integrate third-party services. See the API docs for more details. (by @valentinfrlch)
• Event classification: Events are now classified into categories and labels. For custom events, these can be assigned manually using the create_event action. For events that are automatically added through an analyzer or the blueprint, category and label is assigned automatically. For a full list of all available categories and labels, see the docs. (by @valentinfrlch)

🔧 Improvements & Bug Fixes

• Create Events: The create_events action now supports setting the category and (new) label for an event manually. (Remember Icon #536) (by @valentinfrlch)
• Request Timeout: The request timeout is now configurable (useful for local LLMs that may need more time to respond). (by @ljmerza)
• Timeline: Fixed an issue that caused images to be removed before the event was created, resulting in broken notifications and timeline cards. (Thumbnail disapears once the notification has been updated to include the LLM description #551) (by @valentinfrlch)
• Special Characters: Improved handling of special characters in event titles (by @radzio)
• Unavailable Cameras: Added checks to detect if a camera is offline. Only raises an error if all cameras are unavailable. (by @rmaher001)
• Unit Test Coverage: Added unit tests. (by @Smiley73)
• Fixed a bad request issue during inital setup. (v1.6.0-alpha.1 - Bad Request #574) (by @valentinfrlch)
• Fixed an issue that caused validation to fail for gpt-5 models when using Azure (Adding Azure as provider for OpenAI GPT-5 results in an error. Unable to add as provider. #494) (by @valentinfrlch)
• Fixed an issue that caused newer claude models to fail silently (Anthropic provider sends both temperature and top_p parameters causing API errors with newer Claude models #562) (by @dcgrove)
• Pinned boto3 to 1.37.1 to avoid conflicts with other integrations (boto3 version conflict #580) (by @valentinfrlch)

🌐 Languages

• Support for Hungarian language: Added 🇧🇬 Bulgarian translations. (by @boian-ivanov)
• Support for Turkish language: Added 🇹🇷 Turkish translations. (by @muminkoykiran)

Blueprint

If you're having this issue on an existing blueprint, set this new option explicitly so it changes from the unknown {{ camera }} variable to a static "LLM Vision Snapshot" group.

🔧 Improvements & Bug Fixes

• Importance Classification: The system prompt to classify important events is now configurable. (by @TheRealFalseReality)
• Fixes the “camera is undefined” error. (by @TheRealFalseReality)

---

This discussion was created from the release Structured Output & New Timeline API.