arXiv cs.AIThursday · May 28, 2026FREE

A Query Engine for the Agents

Hyperparam, detailed in arXiv paper 2605.27785, addresses the challenge of querying unstructured text in AI applications. The fastest-growing data in production—agent traces, chat logs, reasoning chains, model outputs—cannot be analyzed with SQL alone because text requires a model in the query path. Hyperparam provides three open-source JavaScript libraries: Hyparquet, Squirreling, and others (names not fully disclosed). These libraries are JS-native, small enough to ship inside a cold tab or per-turn agent sandbox, and interleave analytic operators with model-based text interpretation. This allows AI applications like Claude Code, Cursor, and Claude Desktop to analyze data client-side without relying on Spark, Trino, or managed warehouses. The paper was published on arXiv on May 28, 2026, under the category cs.AI. The libraries are open-source, but specific license details are not mentioned in the excerpt.