Vector search arch


Is there a location where one could learn more about the architecture of the vector DB?
As in how the vectors are stored/how they are indexed, and how mongoDB (the app) comes into play?

If I understand correctly, while the embeddings themselves are stored in mongoDB like normal documents, the indexing for fast searches is done outside of mongo (e.g. apache lucene), is that correct? would love to learn more,


Hello @Roey_Maor

The architecture is somewhat similar to Atlas Search here: . Though there are some differences.

Vectors are stored in your MongoDB documents as a BSON Arrays and once you have defined a “Vector Search Index” on that field, the collection is monitored automatically to build and update the index based on new documents being created or documents being removed from your collection.

Is there a specific question or concern you have around the service?


1 Like

Hello Benjamin,

Thank you.
There really isn’t any specific question I have about the service, was just curious to learn more about how it works under the hood. The link you provided is helpful.