This topic describes how to build typical vector search applications with OceanBase Database.
Vector search tutorials
Build an intelligent document Q&A assistant with OceanBase
The document Q&A assistant is an application built on OceanBase Database. It stores documents in bulk as vectors. Users ask questions through the UI; the application embeds each question into a vector using the BGE-M3 model and searches the database for similar vectors. When it finds the document content for those similar vectors, it sends that content together with the user's question to an LLM, which generates more accurate answers from the provided documents.
Build an image-search application with OceanBase
The image-search application stores an image library as vectors in the database. Users upload a query image in the UI; the application converts it to a vector and searches the database for similar vectors. It then displays the matching images on the UI.
Build a multi-model fusion application with OceanBase
OceanBase's multi-model fusion support enables applications such as a travel assistant that work with multiple data types at once:
- Vector data: Semantic information such as attraction types and food preferences
- Spatial data: Location and itinerary range
- Relational data: Structured information such as attraction ratings and travel season
Multi-model hybrid search over these data types delivers accurate, personalized, and real-time recommendations.