The pg-jev extension has been released for filtering data in PostgreSQL through AI

Developers presented pg-jev — an extension for the PostgreSQL database that allows performing semantic search, classification, and ranking of strings using natural language. The tool integrates the capabilities of the Jev model directly into SQL queries, ensuring data processing without the need to upload to external systems. For the extension to work, the plpython3u interpreter must be installed, which provides interaction between the SQL environment and Python scripts. The user must provide an API key to access the computational power of the model. The filtering process occurs by passing a text query, which is converted into logical conditions for selecting data from tables. The functionality of pg-jev is focused on tasks where fuzzy search or complex categorization of records is required, which are difficult to implement using standard SQL tools. The tool allows using natural language to form filtering criteria, which reduces the time spent writing complex regular expressions or multi-level WHERE conditions. The solution expands the capabilities of working with unstructured data within relational databases. The use of plpython3u makes the tool available for deployment in standard PostgreSQL environments where the execution of Python procedures is allowed. This simplifies the introduction of AI functions into existing backend systems without changing the data storage architecture.