Against national digital-driven educational strategies, college English for non-English majors is troubled by fragmented resources, homogeneous teaching and limited personalized feedback. Targeting first-year non-English students, this study develops a knowledge-graph-enabled smart curriculum for General English. Combining top-down and bottom-up strategies, a two-tier knowledge graph is established: an ontology knowledge layer sorted from syllabus-centered five-skill knowledge points, and a data-resource layer built upon MOOC and SPOC resources on the Chaoxing Fanya platform. Based on a “teacher-student-machine” interactive framework, this paper presents the full-cycle teaching workflow covering pre-class guidance, in-class interaction and post-class adaptive push. Integrating ideological-political education and power-industry materials, this two-year research uses case studies. Pilot results show knowledge-graph visualization clarifies implicit links between language knowledge points, supports adaptive recommendation and multi-dimensional formative assessment. It offers feasible references for digital transformation of college foreign-language courses.