I am shocked how fast it was to implement semantic search on my work project. I still do not fully get it, but am embracing a bit of magic... am waiting for the... indexer, prefer the name, to embed all the data from about a million products using Ollama on the host to create the embedding for me.
we may even self host it, but the code is solid for swapping providers as needed ofc. I just cannot test it as my laptop will be working over night to embed all the things :D
commit 1e42b69
A neat trick is to look into the batch mode if you wanna swap to openai in the future, cheaper, bulk pricing, but not "instant" responses.
Make sure to also look into rank fusion and reranking. Lmk if you want to talk about this