Semantic search using text embeddings. With Open Distro for Elasticsearch you can store your text embeddings and then perform a nearest-neighbor search[1] to find most similar documents using cosine similarity[2]. Elasticsearch (vanilla) will get this feature with 8.0.

If migrating to ES makes you groan you can use a managed service like Pinecone[3] (disclaimer: I work there) just for storing and searching through text embeddings in-memory through an API while keeping the rest of your data in PG.

[1] Nearest-neighbor searches in Open Distro: https://opendistro.github.io/for-elasticsearch-docs/docs/knn...

[2] More on how semantic similarity is measured: https://www.pinecone.io/learn/semantic-search/

[3] https://www.pinecone.io