[Question]: How to put BGEM3Index embeddings to mongo atlas vector store

[takan55]

Question Validation

• [X] I have searched both the documentation and discord for an answer.

Question

How can I put BGEM3Index embeddings to mongo atlas vector store? If possible, code samples will be very much appreciated. thanks

[dosubot[bot]]

Hey @takan55! 👋 I'm here to help you with any coding conundrums you may have. Let's tackle this challenge together!

To store BGEM3Index embeddings in a MongoDB Atlas vector store, you can follow these steps. Below is a code sample that demonstrates how to achieve this:

1. Install Required Libraries: Ensure you have the necessary libraries installed:

   pip install pymongo
   pip install llama-index
   pip install llama-index-vector-stores-mongodb

2. Connect to MongoDB Atlas: Use pymongo to connect to your MongoDB Atlas cluster:

   import pymongo
   
   def get_mongo_client(mongo_uri):
      """Establish connection to the MongoDB."""
      try:
          client = pymongo.MongoClient(mongo_uri)
          print("Connection to MongoDB successful")
          return client
      except pymongo.errors.ConnectionFailure as e:
          print(f"Connection failed: {e}")
          return None
   
   mongo_uri = "your_mongodb_uri_here"
   mongo_client = get_mongo_client(mongo_uri)
   
   DB_NAME = "your_database_name"
   COLLECTION_NAME = "your_collection_name"
   
   db = mongo_client[DB_NAME]
   collection = db[COLLECTION_NAME]

3. Store BGEM3Index Embeddings: Assuming you have the embeddings as a list of floating-point numbers, you can store them in MongoDB Atlas:

   from llama_index.vector_stores.mongodb.base import MongoDBAtlasVectorSearch
   
   # Example embedding
   embedding = [
      0.017273761332035065, 0.0003562009078450501, 0.008221239782869816,
      -0.017780985683202744, -0.04745353385806084, -0.008362134918570518,
      # ... (rest of the embedding values)
   ]
   
   # Document to insert
   document = {
      "document_id": "unique_document_id",
      "embedding": embedding,
      "metadata": {
          "title": "Example Document",
          "content": "This is an example document."
      }
   }
   
   # Insert the document into the collection
   collection.insert_one(document)
   print("Document inserted successfully")

4. Querying the Vector Store: You can query the vector store to retrieve similar embeddings:

   query_embedding = [
      0.017273761332035065, 0.0003562009078450501, 0.008221239782869816,
      # ... (rest of the query embedding values)
   ]
   
   # Example query to find similar embeddings
   results = collection.find({
      "embedding": {
          "$near": {
              "$geometry": {"type": "Point", "coordinates": query_embedding},
              "$maxDistance": 1000  # Adjust the distance as needed
          }
      }
   })
   
   for result in results:
      print(result)

This code demonstrates how to connect to MongoDB Atlas, store BGEM3Index embeddings, and query the vector store for similar embeddings. Make sure to replace placeholders like "your_mongodb_uri_here", "your_database_name", and "your_collection_name" with your actual MongoDB Atlas details ^[1][2].

To continue talking to Dosu, mention @dosu.