Cardiovascular diseases remain the leading cause of mortality worldwide, accounting for approximately 17.9 million deaths annually. Many acute cardiac conditions, including myocardial infarction and severe arrhythmias, are highly time-sensitive, requiring rapid diagnosis and intervention. In clinical practice, cardiology triage involves integrating heterogeneous data sources such as patient history, symptoms, vital signs, electrocardiograms (ECG), echocardiography, and laboratory results. However, these data are often fragmented across systems, increasing cognitive burden and delaying decision-making. We propose a multimodal decision-support framework for cardiology triage that integrates structured patient data, diagnostic examinations, and learning-based test selection within a unified system. The platform supports ECG, echocardiography, and laboratory inputs, which are processed independently and combined through a fusion mechanism to generate an interpretable summary for clinician review. Diagnostic test selection is formulated as a sequential decision process, enabling context-aware recommendations based on the evolving patient state. The system maintains a clinician-in-the-loop design, ensuring that all outputs remain advisory. The framework was evaluated using both publicly available datasets and a retrospective clinical dataset of approximately 500 ECG recordings collected in a cardiology practice and annotated by a board-certified cardiologist. On the clinical dataset, the system achieved accuracy above 95%, sensitivity above 93%, and F1-score above 90%, indicating strong agreement with expert interpretation. This work demonstrates the feasibility of combining multimodal data integration, ECG analysis, and structured decision support within a single system, providing a foundation for clinically relevant AI-assisted cardiology workflows.