October 2025 Summaries

PyDough Community Edition (PyDough-CE) is an open-source project designed to make natural language analytics practical and secure by using a domain-specific language (DSL) called PyDough. Rather than allowing language models to generate raw SQL, which can be risky and untrustworthy due to issues like prompt injection and SQL dialect sprawl, PyDough lets models generate short, auditable plans that are compiled into correct SQL for any database backend. This approach enhances trustworthiness, portability, and auditability by acting as a protective layer between natural language inputs and SQL output, ensuring that generated code is safe and valid. PyDough-CE provides a comprehensive workflow that includes converting database schemas into a knowledge graph, translating natural queries into PyDough plans, compiling these into SQL, and executing the queries, all while maintaining transparency and safety. It is particularly useful for teams managing multiple database dialects, as it simplifies query management and minimizes risks associated with open-ended SQL generation, while its open-source nature allows for community contributions and feedback.

AI agents are increasingly used to automate data-driven workflows, but efficiently managing large datasets remains a challenge due to the limitations of popular libraries like Pandas. Bodo DataFrames addresses this issue by providing distributed execution and HPC-grade performance to standard Pandas code without requiring significant refactoring. By integrating Bodo with LangChain, AI agents can handle large-scale datasets, such as the NYC Taxi's billion-trip dataset, with much greater efficiency. For instance, using Bodo, an AI agent can swiftly analyze and answer complex queries about real-world data, such as estimating taxi fares between specific locations, which would otherwise be impractical with standard Pandas on smaller systems. This approach allows AI agents to scale their operations, leveraging Bodo's compatibility with Pandas and its MPI-based parallel backend to process large datasets without running into memory errors. The example demonstrates Bodo's effectiveness by completing a complex data query in just 4.5 minutes on a 2024 MacBook Pro.