Abstract: Hepatic fibrosis is a critical pathological link in the progression of various chronic liver diseases to cirrhosis and hepatocellular carcinoma, yet no specifically approved anti-fibrotic drug is currently available. This study aims to construct a literature-augmented knowledge graph and integrate multiple knowledge graph embedding (KGE) models to systematically screen drug repositioning candidates with potential anti-hepatic fibrosis activity, and to validate their binding feasibility through molecular docking. We integrated the DRKG hepatic fibrosis subgraph with PubMed literature abstracts, and employed the DeepSeek large language model API to extract SPO triples. After entity standardization, relation normalization, and node reclassification, a knowledge graph comprising 124 327 nodes and 292 533 edges was constructed. Four embedding models—TransE, RotatE, RGCN, and CompGCN—were adopted for link prediction training, followed by candidate ranking integrated with multi-model Z-score normalized ensemble scoring, entity type constraints, and fine-grained classification strategies. Top-ranked candidate molecules underwent evidence-level validation and molecular docking simulations. The ensemble KGE model achieved a predictive performance of AUC-ROC 0.9487, ultimately identifying 5 041 candidate molecules. The Top-10 candidates were predominantly natural products and already-marketed drugs, including succinate, quercetin, and puerarin, all of which were supported by L1_direct-level (candidate molecules have direct therapeutic edges to the target disease node of liver fibrosis in the training graph) evidence. All 14 preset positive controls were detected (100% recall rate), with 14 entering the main ranking list. Molecular docking results showed that the best binding energy of naringenin (Top-10) with FAK (PDB: 2J0L) was −7.28 kcal/mol (1 kcal=4.184 kJ), and that of apigenin (Top-20) with FAK (PDB: 2J0L) was −7.51 kcal/mol, both below −5.0 kcal/mol threshold. Non-bonded interaction analysis revealed key binding driving forces including hydrogen bonds and hydrophobic interactions. This study provides a comprehensive computational strategy for hepatic fibrosis drug repositioning, spanning from literature knowledge integration and multi-model candidate ranking to structural biology validation. The identified candidate molecules, particularly naringenin and apigenin, demonstrate favorable therapeutic potential and offer clear data directions for subsequent experimental validation.