When talking about Vector Databases, in the market we can find the specialized ones and multi-model, most of the major database providers like Oracle, PostgreSQL or MongoDB, for mention some of them, have integrated a specific solution to retrieve vector data.
The key concept is Retrieval Augmented Generation (RAG) and combined with Large Language Models (LLM), we can use models with data that is always changing. It is a use case to deploy a model trained with static data, for example the history of a store sales, but when the data is constantly changing you can give an external knowledge database to improve the response of the LLM.
DB-Engines offers a complete index for databases based on their current popularity, mentions in social networks, frequency of search in Google Trends, frequency of discussion in technical foros and number of job offers, in which is mentioned.
Why a Vector Database is needed ?
The way the data is used for LLM's is in a vector representation. That means the prompt requested by the user goes to the vector database and search for the document with best similarity and return the results ranks to the LLM and personalize the prompt response.
Which service can be used on AWS ?
- MongoDB from the marketplace or directly from Atlas Portal.
- RDS PostgreSQL with the extension pgvector.
- Open Search with the new feature Open Search Serverless.
I am going to focus on pgvector and Open Search Services, this services are infrastructureless oriented, we don't need to care about a lot in administrative tasks with the infrastructure.
Check this blog post Building AI-powered search in PostgreSQL using Amazon SageMaker and pgvector, if you want to use the extension inside RDS for PostgreSQL.
And if you are looking to go Serverless, check this blog post Vector engine for Amazon OpenSearch Serverless is now available
Top comments (1)
I am from SingleStore database. Let me know if you like to know more about SingleStore database