GrapheneOS includes a fork of the open source TalkBack. You need a text-to-speech implementation such as Google's Speech Recognition & Synthesis, RHVoice or eSpeak NG too. None of those is suitable for inclusion in GrapheneOS due to licensing and other reasons. There are some promising new apps based on more modern approaches which we're keeping an eye on. We may make our own text-to-speech and speech-to-text similarly. We recently built our own network location implementation in the OS and are building our own network location database/service with full offline support. The same thing is probably needed for TTS and the other direction.
You can install Google's Android Accessibility Suite for the full proprietary feature set including extended TalkBack features. It works fine on GrapheneOS. It needs sandboxed Google Play for the full feature set.
Why build your own instead of working with BeaconDB? Or if you predate that project, vice versa: will this be an open dataset that other projects can use and contribute back to?
I tried looking up more info about it but DDG only comes up with forum threads about workarounds