Precision over prediction.
Most astrological software treats interpretation as a lookup table — pre-written paragraphs indexed by planet-sign combinations. Zenit takes a different approach. We built a knowledge graph encoding the structural relationships of astrological tradition: how planets express through signs, how houses contextualize, how aspects modify. The AI traverses this graph to construct interpretations specific to each chart.
The result is not fortune-telling. It is a computational tool for exploring the symbolic language of astrology with the same rigor you would apply to any knowledge system.
How it works.
Astronomical Computation
Swiss Ephemeris (v2.10) for sub-arcsecond planetary positions. Placidus and Whole Sign house systems. Full aspect detection with configurable orbs.
Knowledge Graph
12,000+ structured rules encoding astrological tradition. Vector embeddings for semantic search. PostgreSQL with pgvector for efficient retrieval.
Interface
SvelteKit with server-side rendering. Interactive SVG chart visualization. Real-time transit timeline. Designed for data density without clutter.
AI Engine
LangGraph agent architecture. Context-aware interpretation grounded in natal chart data and active transits. Knowledge graph retrieval, not generic language model output.
Where the data comes from.
Every calculation in Zenit is traceable to a specific data source. No black boxes.
Planetary positions derived from NASA JPL Development Ephemeris (DE441). Sub-arcsecond accuracy for dates between 13201 BCE and 17191 CE.
Birth place resolution via OpenStreetMap Nominatim. Timezone detection for accurate local sidereal time calculation.
Interpretations synthesized from classical and modern astrological texts. Cross-referenced knowledge graph with 12+ source works.
Built by
Software Engineer
Zenit is designed and engineered by Diego Alvarez — a software engineer based in San Pedro de Atacama, Chile, under some of the clearest skies on Earth. Interested in the intersection of structured knowledge systems and computational tools.