Troubleshooting detection in intelligent customer service systems is a laborious task. Addressing the challenges of high workload and costs for service representatives in existing systems, this paper introduces a new method for troubleshooting detection within an intelligent customer service system based on the Vertical Tuning and Retrieval Framework (VTRF). The proposed method aims to swiftly provide solutions aligned with human preferences through automation and intelligence. At its core, the method involves constructing a framework comprising four modules. The VertiTuning Module injects domain knowledge into the model to enhance response accuracy and relevance, while the Vector Retrieval Module incorporates real-time updated information to further bolster the model’s response generation capabilities. Finally, detailed experiments are conducted to demonstrate the advantages and potential applications of this framework in troubleshooting detection within intelligent customer service systems.