Aug-1-2026
Add users permissions: Read. Write, Admin
Depending on the user’s role, the agent can safely execute:
Read – SELECT
Write – INSERT, UPDATE
Admin – DDL operations such as CREATE, ALTER, and DROP
Building this project taught me that AI agents are much more than prompt engineering. The real engineering challenges are orchestration, security, permissions, execution flow, state management, validation, and making the system reliable when the model doesn’t behave exactly as expected.
This project has given me a much deeper appreciation for the engineering behind enterprise AI platforms such as Snowflake Cortex AI, Microsoft Fabric AI, and similar technologies.
It’s still a demo project, but it has already been an amazing learning experience. My next goal is to extend it with multiple data sources, MCP support, semantic models, and more advanced agent capabilities.
