Ask HN: Which Vector Database do you recommend for LLM applications?

I have been using FAISS for running https://drpawd.com. It is pretty solid, the only downside is that it's all in-memory so as your dataset grows you need to throw more memory at it.

Qdrant Team here. It depends on you use case. For smaller data amount and just vector similarity search you can use libraries like FAISS, ScaNN, etc. If you are looking for a simple solution without high requirements on performance and just maybe a few millions of vectors you can use traditional solutions like Elastic or Postgres. Dedicated vector db solutions like Qdrant become relevant when you are looking for real-time performance, scalability up to billions and dedicated features that make your life easier. Qdrant works standalone in local (in memory), docker and (managed) cloud mode by just exchanging the host url. Best available performance https://qdrant.tech/benchmarks/ Easy to start with https://qdrant.tech/documentation/quick-start/ And there is a managed Cloud with 1GB free plan. https://cloud.qdrant.ok You are welcome to join our vibrant community on Discord http://qdrant.to/discord

It the oldest project, correct. Regarding popularity, 20K GitHub stars in >6 years, Qdrant 11K stars in https://github.com/topics/vector-database Written entirely in Rust, zero legacy.

ClickHouse if you need search + analytics and full SQL support: https://clickhouse.com/blog/vector-search-clickhouse-p2

Have you tried Marqo? check the repo : https://github.com/marqo-ai/marqo