# Any free vector databases recommendation

**URL:** <https://community.openai.com/t/any-free-vector-databases-recommendation/222019>\
**Category:** API\
**Tags:** api\
**Created:** [May 22, 2023, 1:46am UTC](https://community.openai.com/t/any-free-vector-databases-recommendation/222019 "2023-05-22T01:46:53Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![andrewzheng1618](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/andrewzheng1618/32/134675_2.png) [@andrewzheng1618](https://community.openai.com/u/andrewzheng1618)\
**Post date:** [May 22, 2023, 1:46am UTC](https://community.openai.com/t/any-free-vector-databases-recommendation/222019/1 "2023-05-22T01:46:54Z")

</div>

For less than a million of embeddings and low QPS use case, any free vector database is easy to use?

---

<div class="post-metadata">

**Author:** ![aikanarov](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/aikanarov/32/48170_2.png) [@aikanarov](https://community.openai.com/u/aikanarov)\
**Post date:** [May 22, 2023, 1:48am UTC](https://community.openai.com/t/any-free-vector-databases-recommendation/222019/2 "2023-05-22T01:48:12Z")

</div>

There are a lot, top of my mind, Pinecode, Qdrant… it is easier if you user a hi level like langchain to orquestarte all.

---

<div class="post-metadata">

**Author:** ![jwatte](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/jwatte/32/127705_2.png) [@jwatte](https://community.openai.com/u/jwatte)\
**Post date:** [May 22, 2023, 3:35am UTC](https://community.openai.com/t/any-free-vector-databases-recommendation/222019/3 "2023-05-22T03:35:52Z")

</div>

Do you want to self host?  
[FAISS](https://github.com/facebookresearch/faiss) is good for bigger databases (millions/billions.)  
[bfes](https://lib.rs/crates/bfes) can do pretty well for smaller databases (thousands.)

Other popular options include  
[pgvector](https://github.com/pgvector/pgvector) if you’re already on Postgres (or use a hosted Postgres with plugin support) as it’s better than FAISS for databases that mutate.  
[Redis vector similarity](https://redis.io/docs/stack/search/reference/vectors/) has come up, but I haven’t benchmarked it. It in turn uses [hnswlib](https://github.com/nmslib/hnswlib).