Considerations for the Scalability of MongoDBAtlasVectorStore

[escon1004]

Currently, unfortunately, MongoDBAtlasVectorStore is encountering issues because it tries to embed all dafa of document

	@Override
	public void add(List<Document> documents) {
		for (Document document : documents) {
			List<Double> embedding = this.embeddingModel.embed(document);
			document.setEmbedding(embedding);
			this.mongoTemplate.save(document, this.config.collectionName);
		}
	}

This causes an issue where, if pre-embedded documents or metadata with Base64-encoded images are present, it leads to exceeding the token limit of the embedding model.

Consequently, the change I want to implement is the following function.
Initially, I intended to achieve this by inheriting from MongoDBAtlasVectorStore and overriding the method

	@Override
	public void add(List<Document> documents) 
	{
		for (Document document : documents) 
		{
			
			if(document.getEmbedding() == null || document.getEmbedding().size() == 0)
			{
				List<Double> embedding = this.embeddingModel.embed(document.getContent());
				document.setEmbedding(embedding);
			}
			
			this.mongoTemplate.save(document, this.config.collectionName);
		}
	}

However, the necessary fields for the method call are set as private, making access difficult.

It would be helpful if they were changed to protected

	protected final MongoTemplate mongoTemplate;

	protected final EmbeddingModel embeddingModel;

	protected final CustomMongoDBVectorStoreConfig config;