With increasing awareness of climate change, there is a growing emphasis on the environmental impact of digital solutions. While numerous tools are available to assess softwareenvironmental footprint post-development, few focus on sustainability during the software design phase. To address this gap, we propose EcoDocSense, a framework that supports engineers to evaluate the sustainability of a software design at design time. Utilizing Large Language Models fine-tuned on a catalog of green software patterns, EcoDocSense analyzes software architecture documents to generate sustainability reports, assessing alignment with green software practices to minimize carbon emissions and recommending improvements. As one of the first frameworks targeting sustainability at the design stage, EcoDoc Sense represents a significant advancement, though opportunities remain for further enhancement. In future we plan to extend EcoDocSense’s applicability to a variety of architectural types and documents as well as to provide the capability to estimate carbon emissions.