Abstract: The application of generative artificial intelligence (GenAI) has expanded the scenarios of library knowledge services, while the circulation of data elements has enriched their content. Exploring the synergistic enpowerment of these two forces to enable multi-source and cross-modal knowledge service in libraries is of great significance for the paradigm transformation of library knowledge services. Starting from the three core elements of library knowledge service, actors, content, and scenarios, this paper attempts to explore the connotation, models and development pathways of multi-source and cross-modal knowledge service in libraries. The findgings indicate that actor collaboration, content digital-intelligent transformation, and scenario personalization are the important connotations, whereas artificial intelligence and data elements are the underlying logic of such service. Multi-agent collaboration, multi-source data-driven mechanism and multi-scenario application form its primary structure. In addition, human-intelligence task division, element circulation and digital-intelligence integration are important pathways to promote the development of multi-source and cross-modal knowledge services in libraries. This study is conducive to optimizing the application of AI in library, improving the service level of multi-source and cross-modal knowledge services, and providing valuable reference for future research.