dsh-tool-stat

When agents process numerical data, they often need to perform aggregation analysis on arrays extracted from CSV or JSON files—such as computing means, percentiles, distributions, or correlation coefficients. However, existing tool chains fall short: single-expression calculators can't handle complex statistics, CSV tools only report structure without aggregation, and model-based mental math is error-prone and hard to verify. The dsh-tool-stat plugin addresses these challenges by offering a zero-dependency, pure-function, and deterministic set of statistical tools designed to efficiently cover everything from descriptive statistics to correlation calculations.

The plugin's core capabilities revolve around four key actions: 'describe' provides a comprehensive data summary including mean, variance, quartiles, and more; 'percentile' supports custom quantile computations; 'frequency' reveals distribution patterns; and 'correlation' calculates Pearson or Spearman coefficients. All algorithms are hand-written—such as Neumaier compensated summation and Welford online variance—to ensure numerical stability, with strict constraints to keep inputs and outputs finite, avoiding NaN or overflow issues. A single function call returns a structured JSON report in milliseconds, entirely without external dependencies or state.

This plugin is ideal for developers, data scientists, or agent builders who need quick and reliable statistical analysis. Whether you're constructing data analytics pipelines, validating model outputs, or handling real-time data streams, dsh-tool-stat delivers reproducible computations that let you focus on business logic rather than low-level implementations. Its deterministic design ensures identical outputs for identical inputs, enhancing trustworthiness and debugging efficiency.