Transparency
Methodology & accuracy
Most astrology tools ask you to trust them. We'd rather show our work: exactly how the engine computes, what it approximates, and how it scores against a chance baseline.
How the chart is computed
- • Deterministic engine, license-clean. Ephemeris via astronomy-engine (MIT, VSOP87-grade) — no Swiss Ephemeris, no GPL/AGPL. The engine computes every value; the AI advisor only narrates those computed facts, never invents them.
- • Ayanāṁśa. Lahiri (Chitrapakṣa), anchored 23.85291° at J2000 with the IAU precession rate (≈ arc-minute accuracy). A traditional Sūrya-Siddhānta (vākya) mode is also available.
- • Sidereal positions. Tropical-of-date minus ayanāṁśa. Rāhu = mean lunar node (Meeus); Ketu = +180°.
- • Ascendant. Numeric horizon method, validated against the Sun-at-sunrise identity across three hemispheres. Houses: whole-sign from the Lagna. Navāṁśa: continuous 3°20′ division.
- • Vimśottari from the Moon's nakshatra (120y, 365.25-day year). Panchāṅga from Sun–Moon elongation; Rāhu Kālam = weekday eighth of daytime; hora = Chaldean order from sunrise.
Known approximations (acceptable for an advisory tool): mean node (not true) for Rāhu/Ketu; single Lahiri ayanāṁśa; timezone offsets via IANA tzdata (not LMT for very old births); geocentric Moon. These can shift a boundary-case nakshatra/pada — documented, not hidden.
Do the timings line up? — lift over chance
We keep a corpus of documented lives with dated events. For each event we check whether the engine activates a relevant significator (natural/chara karaka, relevant-house lord, or a planet occupying/aspecting a relevant house) in the active daśā. Then we score the same events against the wrong chart — the difference is the honest signal.
Read this honestly. This is a transparent heuristic proxy on a small hand-verified sample (A: 3 · AA: 4 · B: 1 Rodden-rated) — not proof of predictive validity, which astrology cannot claim. A hit-rate near the control means chance. We publish the control precisely so the number can't be read as more than it is; we're growing the corpus toward hundreds of verified charts.
By life area
What the AI can and can't do
The advisor is grounded: it narrates only the deterministic engine's computed facts, and it can't compute or invent chart data. Sensitive topics — death-timing, medical/legal/financial directives, and crisis — are handled by a deterministic care-first playbook (no AI), in every supported language. Nothing here is a certainty; it's for guidance and reflection. See our privacy model and terms.