Our methodology

Every chart is calculated using Swiss Ephemeris, the same data source professional astrologers and software like Solar Fire use. Planetary positions are accurate to within an arc-second across our supported date range (1900 – 2100).

House system: We default to Placidus for consistency with most Western astrology. Whole Sign, Equal, and Koch are available in settings. House cusps depend on a known birth time and birth location.

Transits are calculated against your natal chart in real time and we use a 1° orb for major aspects. Forecast dates are exact — no fuzzy windows.

We use the Rider-Waite-Smith deck (78 cards). Draws are cryptographically random via the Node.js crypto random pool, not deterministic by user or seed.

Reversals are optional per user — toggle in settings. Interpretations are written by us; the AI personalizes them to the question's context (love, career, self, spiritual, yes/no) but never generates the base meanings.

We use the Pythagorean system. Life Path, Expression, Soul Urge, Personality, and Birthday numbers are calculated from your full birth name and birthdate.

Master numbers (11, 22, 33) are preserved — we do not reduce them to single digits. Personal year, month, and day cycles use standard reduction rules per accepted modern numerology.

The advisor uses your natal chart, current transits, full numerology profile, recent journal entries, captured life events, and prior conversation history as context for every response. We currently default to OpenAI GPT-4o for language generation, with support for Anthropic Claude where configured.

We store your conversations and journals as embeddings in a private vector database (one namespace per user) so the advisor can recall what's been discussed without re-sending the full transcript every time. You can view and delete these memories under Settings → Your data.

We never train Astrelle models on your data. Your messages may be processed by our model provider to answer your request and stored by our infrastructure providers to operate your account. API model providers do not train on your data under their standard API terms.

• • We don't make deterministic predictions about death, health outcomes, or legal verdicts.
• • We don't produce manipulative compatibility warnings designed to create relationship anxiety.
• • We don't show you advertising or sell your data.
• • We don't sell your birth data, journals, conversations, or generated memories.