Google Assistant Errors

I’ve had Google Assistant configured with Home Assistant for a few years now. To my knowledge I haven’t changed anything recently that would affect this integration. I have Google Assistant configured manually, without Home Assistant Cloud.

I have about 250 entities exposed to Google Assistant. Recently all but just a handful of them started experiencing various errors, with the overwhelming majority of them show as “offline” in the Google Home app. A few multi-layer nested light groups do work for basic on / off, but that’s it. I have lights, vacuums, and climate entities exposed. Attempting to control an entity results in “Something went wrong” in the Google Home app.

Looking in the Google Cloud console I see 2 errors over and over again.

GAL_NOT_FOUND

and

Agent responded empty JSON.

Looking at Google’s troubleshooting documentation, GAL_NOT_FOUND indicates expired access tokens. I re-created the API key for the service account, deleted the account credentials in HA, and unlinked and re-linked the “works with google” app in the Google Home app, yet I’m still getting these errors over and over.

Testing a handful of entities in the Google Home Developer Console passes with flying colors. I can see the devices being manipulated by Google Assistant in HA in realtime.

Once I started trying to fix the Assistant side, the Nest integration also stopped working. Nest would work for about 5 minutes, reporting states and allowing control, then it would fail, complaining about access tokens being invalid.

Bump. I’m getting desperate. The WAF is dropping like a rock

I have 260 entities exposed to Google, but I do have Nabu Casa to ease the setup. That works without a hitch, so the number of exposed entities should not be a problem.

I’d probably check if the certificates are still valid and try the obvious ‘no clue if it helps but try just to be sure’ things like restart HA and the router/modem.

Have you also checked if Google itself or other services such as DuckDNS might have problems in your region?

I ended up nuking the GCP project and device control project and configured everything again from scratch. It’s working now. I think my project was getting throttled / banned.