dsh-units

The dsh-units plugin addresses a critical pain point: large language models (LLMs) often fumble unit conversions. Whether it's the affine math of temperatures, the decimal-vs-binary confusion in data sizes (MB vs. MiB), or reciprocal fuel economy calculations, errors can silently corrupt outputs. This toolbox ensures precision by using pure arithmetic over a static unit table, eliminating guesswork.

At its core, dsh-units offers 17 categories—from length and mass to typography and fuel economy—covering common scenarios like cooking units, power definitions, and CSS measurements. It runs with zero runtime dependencies, requiring no network or filesystem access, which guarantees offline reliability and deterministic results. Each conversion returns structured data, including the original value, both units with symbols and names, the result, applied formula, and category, enabling agents to explain their math transparently.

It's ideal for AI agents built on DeepSeek Harness that need accurate unit conversions, such as those generating reports, travel itineraries, or technical documents. By delegating error-prone math to dsh-units, agents can focus on higher-level tasks while delivering trustworthy, detailed conversions for both professional and everyday use cases.