AstroHitch — Product & Technology

Architecture

Full-stack at release candidate

Every layer is live, tested, and integrated. No greenfield surfaces remain for the public launch.

76,199

Lines of Dart

Zero analyzer errors

47

Screens

39 reusable widgets

51

Services

20+ service classes

4

Platforms

Android · iOS · Web · Win

Module Status

Seven modules — all in the shipping product

Module	Status	Screens	Key Capability	Monetisation
Natal Engine + Charts	✓ Live	8+	Swiss-Ephemeris-grade. 5 house systems. Full aspect set.	All tiers
AI Astrologer (Chat)	✓ Live	4	Dual LLM, 4 chat modes, 600+ interpretation entries, streaming.	Aura tokens
AstroSync	✓ Live	6	Relationship graph, unlimited synastry at Silver+.	Silver / Gold tier
AstroMatch	✓ Live	7	Dating profiles, chart-based matching, Soul Match, messaging.	Silver / Gold tier
AstroMarket	✓ Live	8	Chart-aware product recommendations, affiliate / margin.	Take-rate
AstroPlan + Coach	✓ Live	6+	Cosmic calendar, optimal timing, AI coaching with long-term memory.	Gold tier
AstroKnow Profiling	Roadmap Y1	—	Behavioural + astro profiling. Personalisation engine.	Data + personalisation

AI Architecture

Chart-grounded AI, not generic horoscopes

Dual Provider Architecture

Google Gemini is the primary LLM. An OpenAI-compatible endpoint serves as fallback. Switching is runtime-configurable — no code change required during an incident.

Failover risk: Provider abstraction layer isolates the app from any single LLM supplier outage or policy change.

Interpretation Engine

600+ entry proprietary knowledge base covering planet-in-sign, planet-in-house, aspect interpretations, and transit overlays.

Grounded prompts — LLM receives real chart data, not generic sign data
Four knowledge levels: Beginner → Advanced → Keywords
Localised in English, 5 languages queued
Continuously extensible without model re-training

Streaming Protocol

All responses use dual-phase streaming for perceived-performance optimisation.

            Phase 1: real-time text chunks (word-by-word)

            Phase 2: complete markdown prefixed "FINAL:"

            // Consumer pattern

            StreamBuilder<String>(

              stream: llm.streamChat(prompt),

              builder: (ctx, snap) {

                if (snap.data?.startsWith('FINAL:'))

                  return MarkdownBody(...);

                return Text(snap.data ?? '');

              }

            )

Cost & Safety Controls

Per-tier token caps (Bronze / Silver / Gold)
Aura balance checked before every LLM call
Cost alarms on Firestore + LLM spend
Safety prompts for mental-health, medical, legal topics
Provider-side training disabled where supported

Localisation

Six languages, ready to ship

English (EN) Complete

All UI keys · Interpretation engine · Marketing copy

Spanish (ES) Partial

Primary LATAM market. Completion target Q2.

Portuguese (PT) Partial

Brazil market. Completion target Q2–Q3.

Hindi (HI) Partial

India market — 2nd-largest astrology consumer base globally.

Chinese (ZH) Partial

Traditional Chinese, Southeast Asian diaspora focus.

Japanese (JA) Partial

Japan — premium-paying, astrology-affinity market. Y2 target.