As AI applications proliferate, many teams are adopting dedicated vector databases, but this creates new data silos. A recent post from the KingbaseES team advocates for a multi-model approach that embeds vector capabilities directly into existing relational databases. This reduces data movement and simplifies the architecture, as teams no longer need to synchronize between separate systems. The post highlights how KingbaseES integrates vector storage with SQL, allowing developers to query both structured and unstructured data in a unified way. This trend is gaining traction globally, with major vendors like PostgreSQL and Oracle also adding vector support. For data architects, the key takeaway is to evaluate whether a multi-model database can meet AI workload needs before adding another specialized system to the stack. The approach promises lower operational overhead and better consistency, though it may trade off some specialized performance.
A Chinese database vendor argues for multi-model architectures that integrate vector storage with relational data, reducing the need for separate vector databases.