Pipeline chaining or fallback intent

Feature Requests
1

Right now, if you do not have an intent that handles a sentence it will fallback to _common > responses > errors > no_intent and say “Sorry, I couldn’t understand that”.

Would be great to be able to add an intent to use as a fallback so it can be used in automations.

So for example we could use no_intent fallback to send the sentence to service that could be connected to a privatGPT server and get back the answer to the user.

Not sure how this could be implemented.

I know there is a component that let allows you to generate text responses using OpenAI’s GPT-3 model. So this could also be used as a service there.

This issue is about having the fallback intent, nothing else.

I created an issue about this: Option to add a fallback intent · Issue #1499 · home-assistant/intents · GitHub But I been told it was not the right place for it, and since I see, it is not possible to create feature request issues in core but I have to use this forum for it, I created this feature request.

2

Yes, please!

3

An additional bonus for this idea is regarding CPU efficiency. If the sentence can be handled by HA conversation pipeline, that uses way less CPU time than going to my AutoGPT instances to parse sentences and decide intents/service calls. So I would appreciate it very much if it were possible to try HA assist first, and if that came back no_intent, try AutoGPT for a more full featured but costlier computation to analyse.

4

There is a HACS alternative at the moment: GitHub - m50/ha-fallback-conversation: HomeAssistant Assist Fallback Conversation Agent
But would be great to have this in core.
BTW this is not a fallback for a specific intent, but a fallback for all intents.