The rapid advancement of neural text summarization has largely benefited resource-rich languages, leaving low-resource languages underrepresented due to data scarcity and linguistic complexity. This paper proposes a Transformer-based hybrid framework that effectively generates context-aware abstractive summaries for low-resource languages by integrating the strengths of pre-trained multilingual transformer models with domain-specific hybrid enhancements. The framework leverages semantic attention mechanisms, extractive cues, and knowledge-guided rules to improve content relevance, coherence, and fluency in summaries. To address the limitations of existing approaches such as lack of contextual depth, poor generalizability, and over-reliance on large-scale datasets the proposed system introduces an adaptive learning pipeline tailored to low-resource environments. Experimental evaluations across multiple low-resource language datasets demonstrate significant improvements in ROUGE, BLEU, and METEOR scores compared to state-of-the-art baselines, while also maintaining low computational overhead. Furthermore, qualitative analysis reveals enhanced interpretability and domain adaptability. This research contributes a scalable, efficient, and linguistically inclusive solution to the growing need for automated summarization in underserved linguistic communities.