DuckDB's LATERAL join enables calling APIs for each row in a table without explicit loops, using table functions that return zero or more rows per input. The Vector Gateway Interface (VGI) optimizes these remote API calls by exchanging data as Apache Arrow, allowing functions written in any language to run in separate processes.
DuckDB has introduced support for a new SQL IDE, expanding its tooling ecosystem. The documentation covers installation, client APIs across multiple languages (Python, R, Java, Rust, Go, Node.js), SQL statements and functions, data import/export capabilities, and various extensions for enhanced functionality.
A DuckDB extension enables running LLM queries directly on table rows using SQL, returning typed results from TypeSafe's Jev model. Instead of extracting data and calling APIs separately, users define permitted answer options as SQL criteria, ensuring results are typed by construction rather than validation.
A developer uses DuckDB with the Jev AI model to filter Hacker News submissions based on personal taste preferences. Rather than writing SQL patterns, they express filtering logic as natural language sentences that Jev evaluates as probabilistic scores, enabling flexible content curation without complex pattern matching.
VillageSQL has released a new vsql-duckdb extension that embeds DuckDB inside MySQL, enabling users to run analytical queries on data formats like Parquet and join results with MySQL data without leaving their existing database connection. The extension provides three functions: duckdb_scalar() for single values, duckdb_query() for JSON results, and duckdb_status() for version information.
A new DuckDB extension enables users to classify rows in CSV, Parquet files, or DuckDB tables using Jev. It processes approximately 1,000 rows in 10 seconds, offering faster performance than LLMs and better ergonomics than traditional classifiers for data analysis workflows.