Skip to content
Lumindocs
UseBuildChangelog
Open Lumin

Reference

Composite workflows

Eight server-side recipes that fan out 5-15 sub-calls in parallel and return one synthesized result. Each saves 4-8 round-trips per reading and embeds an expert reading pattern (KSK, K.N. Rao, Sanjay Rath, or cross-system) so the LLM does not need to remember the chain.

Note

Workflows live alongside individual tools. The LLM picks them automatically when the orchestration prompt detects a domain match (matrimonial, career, daily, year-outlook, spiritual, pre-verdict, triple-dasha consensus, kundli match).

Why workflows

Phase 6 added textual orchestration guidance to the system prompt. Phase 7 converts the highest-value recipes into deterministic atomic tools. Three reasons:

  • Latency. One round-trip instead of 5-15.
  • Consistency. Every consumer gets the same recipe in the same order.
  • Skip-resistance. An LLM under context pressure may drop a sub-call. The workflow cannot.

Tier 1: Pre-flight integrity

run_pre_verdict_audit

Bundles boundary-warnings, combust-planets, graha-yuddha, and vargottama in parallel. Returns a single confidence modifier (-30..+20) and band (HIGH / MODERATE / LOW). Run silently before any final verdict to catch borderline charts.

Saves: 4 round-trips. Pattern: E (Strict Pre-Verdict Audit).

Tier 2: Cross-system consensus

run_triple_dasha_consensus

K.N. Rao's signature pattern, runs Vimshottari (event-filtered) + Yogini (36-yr) + Jaimini Chara (sign-based) and finds calendar windows where multiple systems independently flag the same range. Returns convergence level: STRONG (all 3) / PARTIAL (2) / WEAK (1) / DIVERGENT (0).

Saves: 3 round-trips + post-processing. Pattern: B. Source: K.N. Rao, "Finer Techniques of Astrological Predictions".

run_karakamsa_spiritual_lens

Sanjay Rath's soul-purpose recipe. Atmakaraka identification → D9 → Karakamsa → 12th-from-Karakamsa → Ishta Devata + D20 + past-life karmic + Bhrigu Bindu. Returns the canonical Jaimini soul-direction reading.

Saves: 6 round-trips. Pattern: C. Source: Sanjay Rath, "Crux of Vedic Astrology" + Jaimini Sutras.

Tier 3: Domain-complete readings (B2B core)

run_marriage_complete_reading

Marriage analysis bundle for matrimonial-platform B2B. Combines CSL promise verdict + spouse characteristics + Manglik dosha + Upapada Lagna (Jaimini) + vargottama strength multipliers + second-marriage check + DBA timing windows + pre-verdict audit.

Saves: 8 round-trips.

run_career_complete_reading

Career bundle for career-coaching B2B. CSL promise + 8-category career signature + profession nature (industry / sector / employment mode / salary band) + job-change-vs-promotion timing + lagna yoga karaka + Jaimini Amatyakaraka + pre-verdict audit.

Saves: 7 round-trips.

run_kundli_match_complete

The matrimonial flagship. Cross-validates THREE compatibility systems: 36-pt Vedic Ashta Koota + KP-rigorous 6-CSL-factor + Manglik check on both partners + Jaimini Upapada Lagna relationship. Emits a cross-system concordance verdict (STRONG / PARTIAL / DIVERGENT) rather than a single-school score.

Saves: 6 round-trips.

Tier 4: Daily / annual cadences

run_today_complete

Today's personal outlook for daily-content / lifestyle apps. Choghadiya + Hora + Tara Bala + Moon transit + next-auspicious-window list. Returns an overall daily tone (FAVOURABLE / MIXED / CAUTIOUS).

Saves: 5 round-trips per daily refresh.

run_year_outlook_complete

Annual report for yearly-subscription products. 12-month forecast + Tajik Varshaphala (solar return + Muntha) + Tithi Pravesha (soli-lunar return) + 5 key Sahams + Yogini dasha + Ashtottari dasha + Sade Sati phases + eclipse impacts on natal sensitive points. Returns a year tone: GROWTH / CONSOLIDATION / CAUTION / TRANSFORMATION.

Saves: 8 round-trips per annual report.

Calling a workflow

Workflows expose the same MCP tool interface as everything else. Every workflow takes the standard birth-data fields (birth_datetime, latitude, longitude, utc_offset_minutes, ayanamsa) plus its own optional parameters. Output is one structured JSON document, the sub-call results plus a synthesis section with verdict, confidence band, and recommended follow-ups.