At my previous company, we were asked to build an AI for generating reports—one that could chat with users after they asked questions and automatically generate bar charts, line charts, and more. At the time, I lacked experience, so I had to implement the context handling and automatic error correction myself, which was extremely time-consuming and labor-intensive.
But now I have a better approach to share with everyone.
Backend: Spring Boot/Cloud + Hermes + AI API
The backend only needs to maintain the relevant table logic and relationships, along with any complex SQL. It then wraps the connection to Hermes and configures the appropriate AI API for Hermes. This way, context handling, automatic error correction, and agent capabilities are all managed by Hermes. In practice, as long as the model isn’t too poor, the responses are excellent for business scenarios.
For BI applications, I can also recommend DB-GPT. I hope this helps others who work in this area in the future.
