Vector databases store embedding vectors with their metadata, enabling efficient retrieval of documents relevant to queries through various indexing and approximate nearest neighbor (ANN) methods. In the context of [[Natural Language Processing (NLP)]], those lists of floating-point numbers are called [[Vector Embeddings]]. In the context of [[Information Retrieval (IR) System]], data sources like these are also called "indexes." [[Pinecone]] is a very popular native vector database host but is closed source. Click [here](https://youtu.be/klTvEwg3oJ4) for vector db in 3 mins. https://github.com/weaviate/weaviate is an open-source alternative. But ref. [[Information Retrieval (IR) Tech Stack]] [[Data Structures - Scalar vs Vector vs Matrix vs Tensor]] Vector databases can extend LLMs with long-term memory by retrieving historical data; they add context to queries by adding relevant data from documents (input); customize the final response through criticism (output). This technique is called [[Retrieval Augmented Generation (RAG)]].  https://www.pinecone.io/learn/vector-database/ [ChatGPT Retrieval Plugin](https://github.com/openai/chatgpt-retrieval-plugin) is a plugin solution.