With the continuous advancement of internet technology, sentiment analysis has gradually become a part of people's daily lives. However, there is currently a lack of a convenient and rapid emotional analysis model available for the general public to use. This study aims to provide people with a channel to alleviate depressive emotions. When users input a situation requiring comfort, our system will output comforting sentences to alleviate the user's depressive emotions, achieving a comforting effect. To this end, this study has trained a BERT model as well as an LLaMA model. The BERT model is capable of determining whether the user's input sentence requires comfort. Meanwhile, the LLaMA model serves as a generation model for comforting sentences to achieve the comforting effect.