Hi 👋🏻 Has anyone here tried to use vector databases approach to tabular data for recommendation purposes? I am doing an experiment to see if it works well to behave tabular data as embeddings and use facebook faiss to create recommendation. Any examples in this would be appreciated 👍🏽 🙏

We just went through this. It doesn't behave well. We ended up using metadata tags in the vector database & combining vector + keyword search Here's a thread with someone else having a similar problem: https://community.pinecone.io/t/how-to-embed-csv-data-with-multiple-columns-for-querying-in-vector-db/890/9

I will check that out, thanks 🙏

I have worked with langchain for unstructured data and it works fairly well