Natural Language-Based Query Generation for Grafana Dashboards

Currently, an essential aspect of the efficient software development process is the use of AI assistants for writing code. However, the activities of developers are not limited to this, and the process of obtaining information on the performance metrics of services and applications can also be simplified. In this work, an innovative tool was created to translate natural language descriptions into precise queries for the Grafana monitoring system, which will revolutionize user interaction with data. Leveraging the capabilities of modern large language models, it is possible to bridge the gap between human language and the complex syntax of queries. This work describes the process of data collection for domain adaptation, the creation of a system for finding the necessary context for the models, and their further training. Additionally, a system for evaluating the quality of generated queries through their execution was created, and its best configuration forms the basis of a chatbot ready for real-world use.