Vector store in the VantageCloud Lake console is for AWS only.
A vector store is a specialized database designed to store and manage high-dimensional vectors, that is, numerical representations of data like text, images, or user behavior. Rather than rely on exact keyword matches, vector stores use a semantic search by storing data as vectors.
Vectors are an integral foundational component of LLM/NL. They are the numerical representation of human language and allows for comparison between differing phrases to be calculated.
Teradata Vector Store provides the capabilities to interact with vectorized data for generative AI. Specifically, it provides a framework and APIs to:
- Interact with an embedding model to generate embeddings for input data.
- Build an index for efficient retrieval of relevant information.
- Provide algorithms for fast and efficient similarity search.
- Interact with a language model (LLM) to generate natural language response for user questions.
- Offers a new internal UDT based data type Vector for handling vectors.
- Provides easy-to-use REST APIs to interact with it through web applications and other interfaces.
- Provides python-based teradataml interface to interact with it.
Key uses of a vector store:
- Semantic search and retrieval: Search content based on meaning, not just keywords. For example, if you ask a question in natural language, the system can pull relevant documents even if the documents don't contain the exact words in the query.
- AI-assisted content discovery: Generate FAQ or Q&A documents and knowledge-based enrichment content.
- Scalability: Search millions of vectors, making it suitable for large content repositories.
- Easily manage semantic search capabilities: Create, edit, delete, and run administrative tasks easily with vector stores.
You must be a database user to interact with vector store. There are two vector store roles, the vector store admin role creates and modifies vector stores objects and the vector store user role who enables the viewing of the vector store objects.