About attaching vector store from MongoDB

API
1

Does anyone know how I can use a MongoDB Atlas database as a vector store?

There is nothing in the documentation regarding the use of external databases.

2

Example document schema:

json

Copy

{
  "_id": <ObjectId>,
  "content": "An interesting piece of text or metadata",
  "vectorEmbedding": [0.123, 0.456, 0.789, ... ]  // The embedding array
  // any other fields you want...
}

Once the index is ready, you can run vector similarity queries via the $search stage in an aggregation. You’ll want to use the $knnBeta operator (or $vectorSearch if available in your Atlas version).

For example, something like:

js

Copy

db.yourCollection.aggregate([
  {
    $search: {
      index: "yourVectorIndexName",
      knnBeta: {
        vector: [0.111, 0.222, 0.333, ...], // query vector
        path: "vectorEmbedding",
        k: 5  // number of nearest neighbors
      }
    }
  },
  {
    $project: {
      content: 1,
      score: { $meta: "searchScore" }
    }
  }
]);
  • k determines how many similar documents are returned based on the distance metric (often cosine similarity).
  • The $meta: "searchScore" can help you see the similarity ranking.

you will have to do your own research on the best way to index but this is a good start.

3

good for you my good brother
I have api at good rate,my t-g:ynnmmoo

4

You might find this useful:

5

langchain_mongodb simplifies a lot

1 Like
6

Yeah could go that route as well. There are lots of options depending on your abilities or if you want an easier drop in with out all the extra code. I like to do my own stuff when. I can :grin:

7

And that is good approach. :v:It is always better to get the understanding of what is going on under the hood.