Currently exposing an entity exposes it to all voice assistant pipelines. This isn’t great if you have multiple pipelines for specific areas that have restricted access vs areas with full access.
Some examples include:
- Child’s room
- Guest Rooms
- Using HA for something like a hotel
Would be great if there was an option to have a global list, then be able to further restrict/expose by pipeline.
Current workarounds I know of are:
- if using local assist, not sure
- If using an LLM assistant, you can give it instructions about restrictions, which LLMs seem to love to ignore
- If using an LLM you can only expose global items and then list out allowed entities in it’s context. This doesn’t work for scripts though as they need to be registered as tools. You can hack around it by making a “pass through” script that simply calls a passed script and exposing allowed arguments for this in the context but this could be a bit dangerous.