Lucene as vector store

[serras]

Apache Lucene is a more performant alternative to indexing that our current version based on a Map. In theory it should be able to perform similarity searches without even having to request an Embedding, but unfortunately the current interface requires us to provide a version of search from one such embedding.

[juanpedromoreno]

On another note, I suggest creating separate modules for "optional" integrations instead of adding this dependency to the core, as it may result in many transitive dependencies.

[serras]

On another note, I suggest creating separate modules for "optional" integrations instead of adding this dependency to the core, as it may result in many transitive dependencies.

Makes complete sense. I've split it into a new module langchain4k-lucene.

Maybe we should do the same with the PostgreSQL one...

[raulraja]

👏