Precision Ephemeris

The Swiss Ephemeris is the gold standard of celestial calculation. It compresses and interpolates the JPL DE431 dataset — the same planetary position data NASA publishes for solar system research — into a format optimized for fast, high-precision lookups. It covers every major celestial body from 13,000 BCE to 16,800 CE, and is the de facto standard used by most serious chart calculation tools worldwide.

A single degree in the zodiac is divided into 60 arcminutes, and each arcminute into 60 arcseconds. Our engine calculates positions down to the arcsecond — 1/3600th of a degree.

The four-minute rule: Because the Earth rotates one degree every four minutes, even a slight error in birth time or a low-precision calculation can shift the Ascendant or house cusps entirely. Arcsecond precision ensures the chart we generate is a faithful mathematical reflection of the sky at the moment of birth.

Many apps rely on the average motion of planets, ignoring the complex gravitational interactions of the solar system. StellarTies accounts for the actual perturbations:

Retrograde stations: We pinpoint the exact moment a planet appears to stand still and reverse direction, which matters for transit timing.

True vs. Mean Nodes: We calculate the True Lunar Nodes, accounting for the oscillations in the Moon's orbit. This provides a more accurate nodal axis than the smoothed "Mean Node" used by many apps.

Chiron and outer bodies: We include Chiron and the True North Node alongside the ten classical planets, with the same precision.

We solve the "hallucination" problem by strictly separating data from narrative:

The math layer: Our calculation engine calls the Swiss Ephemeris library directly. The resulting planetary positions are immutable and mathematically verifiable.

The logic layer: Aspects are detected using exact angular geometry — conjunction at 0°, trine at 120°, square at 90° — with configurable orb tolerances. No interpretation happens here.

The narrative layer: Only after the math and logic are verified does our language model synthesize the data into human-readable insights. The model cannot invent aspects that don't exist in the data.