This topic describes how to build vector search applications on OceanBase Database.
Build vector search applications
Build an intelligent Q&A bot with OceanBase
The document assistant is an application built on OceanBase that stores documents in batches as vectors. Users can ask questions through a user-friendly interface, and the system uses the BGE-M3 model to convert their queries into vectors. It then searches the database for similar vectors. When relevant document content is found, both the retrieved information and the user’s original question are sent to an LLM, which generates a more accurate answer based on the provided documents.
Build an image search application with OceanBase
The image search application is another solution powered by OceanBase, where image libraries are stored in the database as vectors. Users can upload images through the interface to search for visually similar content. The application converts the uploaded image into a vector and searches for similar vectors within the database. The results are then displayed as images directly on the UI, making it easy for users to find related visuals.
Build a multimodal fusion application with OceanBase
OceanBase's multimodal fusion technology supports applications like a travel assistant, which can process multiple types of data at once, including:
- Vector data: Semantic information such as types of attractions or food preferences
- Spatial data: Geographic locations and travel ranges
- Relational data: Structured information like attraction ratings and travel seasons
By efficiently combining and searching across these different data types, OceanBase enables more accurate, personalized, and real-time recommendations for users.