How Ishvaram computes every Vedic fact

The astronomical engine

Every astronomical fact on Ishvaram — sunrise, sunset, tithi, nakshatra, yoga, karana, planetary longitudes, Rahu Kaal, choghadiya and muhurat windows — is computed with the Swiss Ephemeris. Nothing is looked up from a static table or recalled from an AI's memory; the positions are calculated for your exact coordinates and timezone at the moment in question. The engine itself contains no language model and makes no network calls — it is arithmetic over an ephemeris, and nothing else.

The exact parameters we compute with

Most astrology sites publish results without publishing their settings, which makes the results impossible to check. These are ours. They are generated directly from the engine's own configuration, and an automated check blocks any release where this table and the running code disagree — so what you read here is what actually ran.

Reproduce our numbers

You do not have to take any of this on trust. With the parameters above, a birth moment and its coordinates, any competent Swiss Ephemeris implementation should land on the same sidereal longitudes we do — that is the point of publishing them. For a daily check that needs no code at all, pick tonight's sunrise, this week's tithi transitions, or a wedding muhurat, and compare what these pages show with whichever published Panchang your family already follows. Where a small difference appears it almost always traces to a different ayanamsa choice or a neighbouring city's horizon, both listed above. Anything larger than that is a bug we want reported, not defended.

Interpretation is anchored to the shastras

The engine supplies exact positions; what those positions mean is read from the classical texts — Brihat Parashara Hora Shastra and Phaladeepika— not invented. A graha's effect, a dosha's formation and cancellation, a Vimshottari Dasha period's significance: these follow documented Parashari rules applied to the computed chart. Reasonable astrologers can still weigh the same chart differently — interpretation is a tradition, not a certainty.

The strict line: facts are computed, prose is written

Ishvaram never lets a language model emit a date, time, position, or score as a fact. Those always come from the calculation and are substituted into the writing. That is why a reading holds up when you check it against any accurate Panchang, and why two Ishvaram sessions from the same birth details always agree — while a chatbot that guesses positions from memory can hand two people two different charts. The AI Jyotishi reads the computed chart and frames it in plain language; it is disclosed as an AI and never a fabricated human.

Your birth place, timezone, and the local horizon

A kundali is only as trustworthy as the birth moment behind it. When you enter a birth place, Ishvaram resolves it once to precise latitude, longitude, and its IANA timezone — including historical clock changes — and a purchased reading keeps that resolved location frozen with the order, so a chart can never silently shift because a city was looked up differently later. Every displayed time passes through one timezone-aware formatter; nothing is ever shown as bare UTC. Rahu Kaal, choghadiya, and panchang windows are computed from your own city's sunrise and sunset on that specific day, never from a fixed 6-to-6 assumption or a national average clock.

The calculation families this covers

The same engine and the same parameters sit under every surface: the free kundali and its divisional charts, Vimshottari dasha and antardasha periods, gochar transits of Shani and Guru, Ashtakoota gun milan with the Mangal, Nadi, and Bhakoot dosh checks, Sade Sati phases, Pitra Dosh and Kaal Sarp Dosh detection, the daily panchang with choghadiya, Rahu Kaal, Yamaganda, Gulika Kaal and abhijit muhurat, Ank Jyotish numerology (mulank, bhagyank, name number), nakshatra-and-pada based baby name syllables, and prashna kundali. If two pages ever show the same fact, it came from one computation — there is no second copy to drift.

How calculations are calibrated and verified

Calculation surfaces carry automated calibration tests that pin the engine's output against independently published traditional Panchang references. The tolerances are strict and are asserted on every build: sunrise and sunset within ±5 minutes, Rahu Kaal / Yamaganda / Gulika Kaal within ±1 minute, tithi and nakshatra names matched exactly, Gun Milan koota scores matched exactly (the Ashtakoota system is deterministic, so there is no rounding excuse), and festival dates matched to the exact day. A separate drift gate re-compares the engine's own tithi resolution against the published festival dates on every change, so a silent divergence cannot reach a page unnoticed.

This is not a one-time claim: an automated quarterly run re-validates the engine against independently published traditional Panchang references at least every 90 days and fails loudly on any divergence. The most recent cross-validation passed on 1 July 2026 — the full run log is public, so anyone can audit exactly what was checked and when.

Calendar and festival dates

Festival, eclipse, and sankranti dates have a single source of truth in this codebase — there is no second copy to drift — and the ones we publish are pinned by tests against traditional references. Where the engine's tithi resolution and a traditional Panchang genuinely disagree, the traditional date wins and the divergence is recorded rather than papered over: the Amanta and Purnimanta reckonings differ by design, and a date that shifts by a day between regional traditions is a real feature of the calendar, not a defect to hide. Regional variation is expected on tithi-based festivals; where your local tradition differs from what you see here, your tradition is not wrong.

Why two Panchang sources can print different dates

The most common question we get about a festival date is why a family Panchang shows a different day. The answer is almost never the astronomy. Every Panchang computes the same tithi from the same Sun–Moon geometry; what differs is the convention that decides which civil day owns that tithi. Some festivals are bound to the tithi as it stands at sunrise, some at midday (madhyahna), some at dusk (pradosh) — and when a tithi begins or ends across one of those moments, the very same rule lands on two different days.

We measured that effect rather than asserting it. Across the 85 major festivals whose dates our engine resolves without a calibrated override — 170 festival-years over 2026–2027 — 5 are governed by a convention other than sunrise (10 festival-years). Reading the tithi naively at sunrise instead of applying the governing convention changes the day in 6 of those 10 cases, always by exactly one day, and the naive reading always lands on the later day. The remaining festivals are sunrise-governed, so no shift is possible for them by definition.

Measured with the same engine that serves the site, over 2026–2027. “Shifts one day” means the naive sunrise reading of that identical tithi rule falls a day after the date the governing convention gives.

This is also why we publish an open discrepancy rather than hiding it. For Holi 2027 our engine's own resolution falls one day after the date we publish; the published date follows the calibrated cross-source reference, and the unresolved piece is the Bhadra-rahit rule that decides Holika Dahan's pradosh window. That case is tracked as a failing calibration test in our codebase until the rule is implemented properly, because a festival date is the kind of fact people plan a ritual around.

The classical corpus we interpret from

Interpretation rules are drawn from the standard Parashari canon: Brihat Parashara Hora Shastra (the foundational hora text), Phaladeepika (Mantreswara's results treatise), the Ashtakoota matching framework used across North Indian traditions, and the Vimshottari Dashatiming system. Where traditions genuinely disagree — contested attributions exist in the literature, such as per-tithi presiding deities — we deliberately publish only the uncontested layer rather than presenting one lineage's view as universal fact. Reference tables across the site (nakshatra lords, koota matrices, dasha years, divisional charts) are generated from one canonical dataset, so a value can never drift between two pages.

What the AI is — and is not — allowed to do

The AI Jyotishi is disclosed as an AI on every surface where it speaks. It is architecturally prevented from inventing computed facts: dates, times, positions, and scores reach the page only from the calculation engine, and the AI's role is limited to explaining what the computed chart means in plain language, in the vocabulary of classical Jyotish. It never pretends to be a human astrologer, never fabricates credentials, and guidance is framed as tradition and interpretation — a lens for reflection and timing, not a substitute for professional medical, legal, or financial advice.

Reviewed for traditional accuracy

Ishvaram's Jyotish methodology is reviewed for traditional accuracy by Pandit Kuldeep Shastri, a practising Vedic astrologer who serves at the Geeta Temple in Karnal, Haryana, with over 26 years of experience in kundali analysis, muhurat, dosha diagnosis, and remedial astrology (upay). The engine computes the facts, the shastras decide their meaning, the AI Jyotishi frames them in plain language, and an experienced human astrologer vouches for the approach.

Our research

The same first-party measurement discipline produces our published research. The first study, What 351,000 conversations with an AI astrologer reveal about India's real worries, analyzes seven weeks of anonymized, category-level usage — what India actually asks, when, and in which language — with its full methodology and privacy notes published alongside the findings. The second study, The Birth-Time Problem: one in ten kundalis is made without a birth time, measures how often seekers lack an exact birth time and what that means for chart accuracy, using the same anonymized first-party discipline. Every published study lives in the Ishvaram research index.

Corrections

If a value here ever disagrees with a careful traditional Panchang, tell us through the contact page. A reported discrepancy is checked against the independent authorities above; a confirmed error is corrected at its single source, and an automated regression test pins the corrected value so the same mistake cannot quietly return. A wrong number under a sacred decision — a marriage date, a naming ceremony, a griha pravesh — is treated as an outage, not a typo: trust is easier to keep than to rebuild.

Citing this page

Researchers, journalists, and astrologers are welcome to reference this methodology, and to quote the parameter table when describing how Ishvaram computes. The stable address for this page is https://ishvaram.com/methodology/. A suggested reference: “Ishvaram, How Ishvaram computes every Vedic fact, reviewed by Pandit Kuldeep Shastri, https://ishvaram.com/methodology/”. If you need a detail that is not published here in order to check our work, ask through the contact page and we will document it on this page rather than answer privately — a methodology that is only explained in email is not a methodology anyone can verify.

See also the About page, our disclaimer on using guidance responsibly, and try it on the free kundali tool.