Awesome repositories on GitHub compile resources in specific fields to support software learning and development. These repositories are widely favored by developers and often contain rich textual descriptions in their README files. However, users often struggle to find appropriate awesome repositories because they typically contain little code and limited metadata. Existing repository recommendation services fail to fully utilize the informative content of README files, which are crucial to the utility of awesome repositories. To address this challenge, we propose ARrec, the first recommendation service specifically designed for awesome repositories using the Graph Neural Network (GNN). ARrec enhances the representation of README files by incorporating external knowledge from Wikipedia and Stack Overflow, enabling the model to better capture semantic similarities between repositories. We evaluate our method on a large dataset of awesome repositories and demonstrate that it significantly outperforms baseline methods, including Large Language Models (LLMs), in recommendation accuracy. Therefore, our recommendation service can assist GitHub users in discovering relevant and high-quality resources aligned with their interests.