AskStar Methodology

The core of AskStar is a three-system chart-computation engine — Western astrology, Zi Wei Dou Shu, and BaZi — not an AI fortune-telling app. The chart itself is a deterministic astronomical and calendrical calculation, fully separable from any AI output.

All three systems are fully deterministic: the same birth input run at any time produces an identical chart. There is no randomness, no model variability, and no AI involvement at this layer.

The sections below name the actual engines and rules each system uses. We only state what maps to code — we would rather claim less than overclaim.

Astronomical computation uses the Swiss Ephemeris (via the pyswisseph library), running in a dedicated Python computation service. The Swiss Ephemeris is the high-precision ephemeris standard used by professional astrology software.

The computation flow: birth date and time are converted to Universal Time (UTC) and a Julian Day, then the ecliptic longitude, zodiac sign, and house are computed for ten bodies — Sun, Moon, Mercury, Venus, Mars, Jupiter, Saturn, Uranus, Neptune, and Pluto.

Houses use the Placidus system: the twelve house cusps, the Ascendant, and the Midheaven (MC) are computed from the birthplace's coordinates. When birth coordinates are missing, only planetary positions are computed — houses are not force-guessed.

Aspects use fixed orbs: conjunction 0° (orb 8°), opposition 180° (8°), trine 120° (8°), square 90° (7°), sextile 60° (6°), and quincunx 150° (3°).

Retrograde status is derived from each body's actual computed longitudinal speed — a negative speed marks the planet retrograde — rather than from lookup-table approximations.

Zi Wei Dou Shu charts are erected with the open-source iztro engine: from the solar (Gregorian) birth date and the traditional two-hour time period (時辰), it establishes the twelve palaces and places the fourteen major stars and the minor stars. Star placement follows the traditional rules as implemented by iztro — we do not modify the star-placement logic ourselves.

Star brightness is normalized to the seven traditional levels: 廟 (temple), 旺 (prosperous), 得地 (favorable), 利 (advantageous), 平 (neutral), 不得地 (unfavorable), and 落陷 (fallen). Abbreviated forms from different sources are normalized to the full level names.

The four transformations (四化 — 化祿, 化權, 化科, 化忌) are computed by the engine and pass through the same normalization layer, so every variant spelling maps to the standard transformation names.

BaZi charts use the open-source lunar-javascript calendrical library: the Gregorian birth moment is converted to the lunisolar calendar, from which the year, month, day, and hour pillars (Heavenly Stems and Earthly Branches of the sexagenary cycle) are derived, along with the zodiac animal and lunar date.

Boundary rules follow the solar-term convention (not lunar month starts): the year pillar changes at the exact instant of 立春 (Beginning of Spring), and the month pillar changes at the exact instants of the twelve 節 terms (立春, 驚蟄, 清明, and so on) — two births on the same day, before and after the term instant, get different month pillars. We verified this behavior directly against the pinned library version.

Leap months are handled explicitly per lunisolar rules: the conversion layer detects leap months (represented as negative month indices) and flags them explicitly in the chart.

AskStar's monthly sky and calendrical data — planetary sign ingresses, lunar phases, solar terms, the month pillar, and zodiac-animal/month-branch interactions — is computed month by month by a dedicated data pipeline and stored in our database. It is not hand-entered and not AI-generated.

Solar terms (節氣) use exact transition instants, expressed in China Standard Time (UTC+8) — the conventional reference clock for solar-term boundaries in Chinese metaphysics. Each year must yield exactly twenty-four terms; one missing or extra term fails the whole computation (fail-loud), and incomplete data is never published.

The month pillar is segmented precisely at the 節 boundary: each Gregorian month contains exactly one month-establishing term, and the pillar changes at that exact instant.

Every computation run is written to the database as a new, incrementally versioned row (insert-only versioning) with full engine-version provenance, so the data is auditable end to end, and a failed run can never overwrite the last good dataset. Pages across the site are progressively being wired to read from this computed dataset.

AI is invoked only after the deterministic computation is complete, and its role is narrow: first, cross-system synthesis — cross-referencing the three independently computed charts (Western, Zi Wei, BaZi) to surface consensus signals and tensions, producing one integrated reading rather than three disjoint reports; second, personalized framing — restating metaphysical terminology in the user's own language, for the user's actual question, context, and report type.

Knowledge-base (wiki) content runs through a pipeline built on the same principle: chart and sky data is computed by the deterministic engines, independently of any AI; AI assists in drafting the interpretive text under anti-fabrication constraints — it is explicitly forbidden from asserting specific calendar dates or planetary transit timings; the founder reviews each page by hand, and only pages that pass review carry the on-page review statement (reviewer name and date).

The review statement is fail-closed by design: the moment a page's content is regenerated, its prior review stamp is voided and removed until the page is re-reviewed. Unreviewed pages show no review claim at all — we would rather show nothing than show a false "reviewed" label.

AskStar's content is for reflection, self-exploration, and entertainment.

It is not medical, psychological, legal, or financial advice and is not a substitute for professional consultation. Decisions about health, money, relationships, or other consequential matters should involve qualified professionals in those domains.

AskStar's interpretations are grounded in traditional metaphysical systems and classical texts. They are not scientifically validated predictions. Users should treat them as one input among many and remain solely responsible for their own decisions.