The growing application of deep learning in architectural design has created opportunities for developing computational and knowledge-driven approaches to sustainable design. However, many existing methods rely primarily on geometric or image-based representations and provide limited mechanisms for integrating climatic knowledge, spatial organization, and building performance requirements into intelligent design processes. Meanwhile, the bioclimatic knowledge embedded in Iranian traditional architecture, particularly in hot-arid climates, has largely remained within historical and qualitative studies and has rarely been translated into computational structures suitable for intelligent design systems. This study addresses this methodological gap by adopting the Design Science Research (DSR) methodology to develop a conceptual and methodological framework for mapping the bioclimatic knowledge of Iranian traditional architecture into contemporary sustainable architectural design. The proposed framework transforms architectural knowledge from qualitative descriptions into structured computational representations through a knowledge base, architectural feature matrices, mapping matrices, and graph-based representations. These components establish explicit relationships among traditional architectural features, climatic functions, design variables, performance indicators, and computational workflows. Within this framework, a conceptual Knowledge-Integrated Graph Generative Network (KIGGN) is proposed to represent spatial, topological-climatic, and functional relationships within knowledge-aware generative deep learning processes. In addition, a conceptual Building Information Semantic Visualization Pipeline (BISVP) is introduced to define the transition from graph-based computational representations to Building Information Modeling (BIM) and Building Energy Modeling (BEM) environments while preserving semantic, spatial, and environmental relationships. The framework also proposes a knowledge-guided loss function incorporating feature reconstruction, spatial coherence, topological validity-validity of structural relationships-, and environmental performance constraints as complementary objectives for future intelligent learning strategies. The primary contribution of the study is not the development or experimental validation of an artificial intelligence model, but the formulation of a structured research artifact that connects traditional architectural knowledge, sustainable design principles, and deep learning methodologies. The proposed framework provides a theoretical and methodological foundation for future computational implementation, empirical validation, and development of intelligent climate-responsive architectural design systems.