This study introduces the Astrological Persona Model (APM), a multi-layered computational framework designed to reformulate archetypal constructs from natal astrology into a dynamic temperament taxonomy for Large Language Models (LLMs). By decoupling astrology from metaphysical claims, the APM utilizes natal data to establish a deterministic baseline (priors) that guides the probabilistic generation of the LLM. The architecture integrates four distinct layers—Natal Core, Qualitative Modulators, Temporal Perturbations, and Sociocultural Filters—to generate a context-aware, stable, and scalable digital persona within an eight-trait latent vector space.Beyond mere simulation, the framework introduces 'Legacy Mode', a paradigm for Voluntary Persona Continuity and Intellectual Succession. This allows individuals to distill their decision-making heuristics and ethical priorities into a query-able 'behavioral weight' system, functionally analogous to applying a Low-Rank Adaptation (LoRA) atop a frozen base model. The study further addresses the ethical imperatives of digital legacy, proposing a Tiered Consent Framework to balance the 'right to be forgotten' with the preservation of cognitive heritage. Experimental results demonstrate that the APM significantly enhances persona consistency and pragmatic realism, providing a mathematically grounded yet empathetic mirror for human-AI interaction.