Automatic Text Summarization (ATS) systems generate short summaries, allowing for efficient extraction of significant information from lengthy text documents. We present an unsupervised extractive ATS approach to Konkani text summarization using distance-augmented sentence graphs, a method that represents sentences as graphs where sentences are connected based on their similarity relationships and the distance between them. Konkani is a language with limited resources, predominantly spoken on the western coast of India, and it possesses a restricted array of language processing tools. Therefore, it benefits from graph-based summarization techniques that perform effectively without extensive training data. Our method involves creating distance-augmented sentence graphs using sentence representations derived from four pre-trained word embedding models: Facebook’s fastText, IIT-Bombay’s Konkani-trained fastText and ELMo, and AI4Bharat’s IndicBERTv2-SS. These representations are subsequently used for calculating the sentence similarity. The system-generated summaries were compared to the reference summaries produced by two separate Konkani language specialists. The performance of each model was compared to baselines. The IIT-Bombay fastText model performed the best, with all of the models surpassing the baselines.