I appreciate your help! Seems like you led me down the right path. 
I took a backup of the table with pg_dump, deleted the old duplicate entry, and the bootstrap completed successfully. 
Thanks for the confirmation! I figured I might need to update the data container (or otherwise get pgvector installed). I was holding off on doing that as I didn’t want to deal with the downtime.
From the discourse-ai thread, it sounds like PG15 is on the horizon, so I might just wait a bit for that.
It seems like the pgvector dependency might be for the Embeddings feature which I didn’t plan on using, but unfortunately it sounds like it’s all bundled together. I mainly just wanted to play with some of the OpenAI / ChatGPT magic rewrite features, so perhaps one more reason to wait until the next big data container update. 