Chinese gaming and social media giant Tencent on Monday officially announced the launch of Tencent Cloud VectorDB, a vector database product for large language models (LLMs) training needs.

Tencent Cloud, the giant's cloud computing arm, noted that the vector database has been applied in more than 30 business scenarios, including QQ Browser, Tencent Video, Tencent Games, QQ Music, and Sogou Input Method. The company also self-developed Olama, the distributed core engine for the vector database.

A vector database is a structured collection of vectors, where each vector represents a data point or an entity.

It plays a crucial role in training LLMs represented by GPT-3.5. Usually, LLMs require vast amounts of data to learn from, and a vector database enables efficient retrieval of similar data points, facilitates faster search operations, and provides a structured and organized representation of entities.

By leveraging vector databases, training can be accelerated as the models can access and process relevant data points more effectively, leading to improved performance and generalization.

Tencent Cloud proposes that vector databases should not only support natural language queries but also deeply integrate AI algorithms into the compute layer, storage layer, and database engine, thus improving the development efficiency of AI native applications.

According to the company, users are allowed to leverage AI capabilities throughout the entire process of utilizing its VectorDB.

Tencent also said its VectorDB is capable of shortening the time required for enterprise users to incorporate LLMs from one month to three days.

Additionally, in the storage layer, the vector database has implemented intelligent compression algorithms, effectively reducing costs by 50%.