I would like to ask the community the best way to store the vector db securely.
I wrote a custom function that encrypts the vector db on the filesystem, but I think making it compliant with lang-chain tools is challenging. Maybe there is an enterprise-grade open-source solution out there that I didn’t find.
Hey, thank you for bringing up this important topic.
While SambaNova is vendor-agnostic regarding vector databases, I can suggest a few tools that integrate well with LangChain: ChromaDB
, Pinecone
, PGVector
, Milvus
These options are quite advanced, but it’s essential to recognize that each enterprise has unique requirements and varying levels of tolerance regarding encryption, privacy, and other security measures. For those considerations, you might also explore: MongoDB
, Elasticsearch
Please note that these suggestions are not formal recommendations from our side; rather, they aim to provide a starting point for your exploration. I encourage you to look beyond this list as you evaluate your options. Thank you!
Thanks, I used ChromaDB as the source. I backup and restore ChromaDB to and from an encrypted file.
I try to make the code as simple but secure as possible. In ChromaDB, I separated the vector database and document store, which allowed me to keep ChromaDB small and fast.
Hi
I am founder of VectorX (https://vectorxdb.ai ). It allows to encrypt the vectors and do fast ANN searches inside the encryption. Will be happy to give you a demo. Please send an email to vineet@launchxlabs.ai
Thanks for letting me know. I started a different part of my project, but I will contact you once I’m back to the vector store!