Orthonym: verified IUPAC names for chemical structures

The 2013 IUPAC recommendations define one preferred IUPAC name (PIN) per structure; naming by hand is slow and error-prone. Most structure-to-name software is proprietary, so a wrong name cannot be traced to its rule. Neural translators can give a valid name for the wrong molecule, and nothing in the name signals the error. Open rule-based tools do not refuse to answer when they cannot verify a name. Orthonym is an open-source, rule-based engine that generates IUPAC names in four stages: perception (RDKit), rules (2013 IUPAC recommendations), assembly and checks. A name must pass three validation checks before it is emitted: atom coverage, a round trip through OPSIN, and an InChIKey match with the input. If any check fails, the engine declines and states the reason. A name counts as correct only if OPSIN reads it back to a structure with the same full InChIKey as the input; a decline counts as a failure. Across HOLDOUT2000, QM9, ChEBI, PubChem (500,000) and ZINC22 (500,000), Orthonym reaches a 97.15–99.98% round-trip rate and emits zero names for the wrong molecule in all test sets, compared with Openclatura 0.3.2 and NISPO 0.1.6. MIT licence. Engine version measured: 1.0.8, available on PyPI and conda-forge (pip install orthonym). Web application: orthonym.decimer.ai