The Core Differentiator
The Confluence Engine
Every other app averages its techniques into a mush. Confluence adjudicates them — a hierarchy of authority where a promise engine can veto, a timing engine leads, and a strength score can only ever dampen, never manufacture a "yes."
The System Stack
A hierarchy of authority, not a democratic vote
The problem with combining Vedic systems is that averaging them lets a weak signal cancel a decisive one. Confluence instead assigns each engine a role — and the role changes by question domain.
Promise (with veto)
The domain's promise authority decides whether the matter is even in the chart. If not, it is a hard veto — no amount of good timing overrides an absent promise.
Timing (the primary voice)
The dasha/KP authority (Vimshottari) sets the "when." Corroborators can only nudge it ±15%; they cannot overrule it.
Trigger
The canon-corrected Jupiter + Saturn double-transit (with retro-station boost) pulls the timing trigger.
Modulators (dampen-only)
Ashtakavarga, Shadbala, combustion, Sade Sati scale the score down, clamped ≤ 1 — they can never originate a vote.
Contradictions (always surfaced)
Honest disagreement — e.g. "timing without promise" — is always computed and shown, never hidden, each with a bounded penalty.
Two-axis audit
Every weighted engine is graded twice: do we trust the code (evidence from tests/lock/parity), and do we trust the astrology (fidelity to BPHS/Jaimini/KP)?
How It Works
How a verdict is adjudicated
- 1
The question is routed to its domain (marriage, career, wealth, health, progeny, litigation… 13 in all), which selects the authoritative engine for each role.
- 2
Promise resolves first — an absent promise hard-caps the whole verdict, no matter what timing says.
- 3
Timing sets the window; the double-transit trigger tests whether it is firing now.
- 4
Readiness and trigger compose non-linearly (readiness^0.65 × trigger^0.35), with a "dasha-dominance rescue" so a strongly promised-and-timed matter is not zeroed just because the rare double transit is quiet.
- 5
Modulators dampen, contradictions are surfaced, and hard caps apply last (promise veto; degraded-ephemeris cap) — producing a confidence with its full reasoning chain shown.
What this is — and what it isn't
- The adjudication core, the 13-domain routing map, and both trust registries are fully built and test-pinned — and already reach users as written guidance through the AI Copilot.
- The polished confidence-dial verdict screen is still a preview surface; it is being wired to run on your live chart, not yet a finished user-facing dashboard.
- A rule-based synthesis of classical Vedic techniques — not a statistically back-tested predictor. It surfaces classical indications, not guaranteed outcomes.
- It weights and adjudicates classical techniques transparently — it does not claim statistical proof that the verdict is correct.