The development of effective diagnostic methodologies for software system failures is of paramount importance. Traditional methods, which rely on specialized terminology and intricate reasoning, require users to have a technical background, resulting in reduced flexibility and decreased user-friendliness. With the rise of generative large language models, optimizing human-computer interaction has become a critical area of focus. Additionally, the inherent intelligence and extensive knowledge of large language models make them both easy and effective to employ for fault diagnosis assistance. We introduce IFKG, an advanced tool for diagnosing software system failures. IFKG integrates generative large language models with knowledge graphs, employing natural language interactions to implement fault detection and deliver solutions. IFKG enables users to upload descriptive problems, retrieve pertinent information from the knowledge graph, and present diagnostic results in natural language. Our accuracy assessments across diverse software system failures indicate that the IFKG provides targeted and actionable recommendations, effectively assisting users in addressing a range of software system issues. The tool is available on GitHub at https://github.com/mako-xx/IFKG, and the demo video can be found on YouTube: https://youtu.be/Die2vgZm2hk.