Calculation standards
The production convention, dependency versions, data snapshots, supported date/latitude limits and tolerance thresholds are documented on the method page. Expected planetary and Ascendant values come from an independent validation environment. The shipped engine must pass the relevant golden fixtures plus boundary, uncertainty and civil-time tests.
A dependency update is a calculation change even when the interface looks identical. Astronomy, timezone-boundary, Temporal or place-data updates require targeted regression tests and a version/change record before publication.
Source and citation standards
Technical claims prefer primary documentation from the library or standard owner. Timezone claims point to IANA and Temporal documentation; accessibility claims point to W3C; privacy and cookie explanations use official public guidance and must still receive jurisdiction-specific legal review. A source link supports the statement but does not transfer editorial responsibility to the source.
Swiss Ephemeris is described precisely as validation-only. GeoNames receives visible CC BY 4.0 attribution. No competitor copy, horoscope article or generated citation is treated as an authoritative calculation source.
Responsible astrology language
The utility may describe a calculated placement and its bounded traditional associations, which are labelled as interpretive rather than factual or deterministic. It must not diagnose, predict certainty, exploit fear or dependency, or give health, legal, financial, employment, fertility or safety advice.
The standing caveat is: Astrology is not scientifically established; interpretations are for entertainment and self-reflection. Placement output itself is not evidence of personality, compatibility or a future event.
AI assistance and human accountability
AI tools have assisted with code, tests, research organization and draft wording during development. Their output is reviewed against the documented requirements, executable behavior, primary sources and automated evidence. AI is not credited as an expert, does not create the independent expected values and does not approve publication.
The publisher remains responsible for approving material content, legal statements, operational behavior and correction handling. Any author biography or expert-review claim must identify a real person and verifiable role.
Corrections, review cadence and change log
A reported calculation issue is reproduced with non-sensitive fixture data, compared against the documented convention and assigned a test before a fix is accepted. A factual or wording correction is checked against its source and neighboring pages. Material changes receive a dated note; silent rewrites are not used to hide an error.
Method and policy pages are reviewed after affected dependencies or behavior change and at least quarterly while the service is active. Calculator copy is reviewed whenever inputs, uncertainty rules, supported ranges or outputs change. The current page-level review record appears below.