Author here. The post here is a 3rd-party translation and may drifts from original wording in a few spots. Azure and OpenAI sent me some corrections today, so I published an updated English version myself here: https://pigsty.io/blog/db/openai-pg/
I've set up a supplementary APT/YUM repository, which builds on the official PGDG offerings. This includes 326 extensions for EL distros and 312 for Deb distros, encompassing 121 RPM packages and 133 DEB packages.
These packages support PG16 extensions across Ubuntu 22.04, Debian 12, EL8, and EL9. My efforts have been particularly focused on align OS-specific extensions across the major Linux distributions to ensure a consistent feature set.
Why not continue to use rpm/apt for building and packaging? You can reuse these packages in the Dockerfile and the image, but not the other way around.
Because of the little gains and large additional effort required in exchange.
rpm/apt are quite involved to package and the tooling is (at least compared to OCI) limited. They would provide little advantage over OCI which works on any packaging system and even OS. They don't enjoy the ecosystem advantages of OCI, despite their age.
In summary, it could be doable, but it would require much more additional effort than packaging in OCI first.
But sure, since it can be done, maybe someone wants to take on it. I won't be down for the job myself ;)
also with pg_trgm[0] (mentioned by OP) and pgvector for semantic search you have a pretty powerful search toolkit. for example, combining them for Hybrid Search [1]
Still very limited and frankly all a bit low level primitives. Unless you are a search expert, you won't be able to do much productive with this stuff. If you are, it might fit a few use cases. But then, why limit yourself to just this stuff?
The point of that of course being that the target audience for this stuff is actually people that for whatever reason are a bit shy using the right tools for the right job here and are probably lacking a lot of expertise. The intersection of people with the expertise that would be happy with this narrow subset of functionality is just not a lot of people.