Possibility of Leveraging Text-to-SQL Capabilities with NeMo Guardrails

[grupocopa]

Hello NVIDIA Developer Community,

Our team is working on a project in the medical domain, aiming to streamline access to information about medications. Specifically, we are building a solution to enable healthcare professionals to quickly query databases of approved medications, their availability, and other critical data using natural language.

We are exploring the use of NVIDIA NeMo Guardrails to implement a Text-to-SQL pipeline that transforms conversational inputs into SQL queries. This would allow us to provide a secure and reliable way to query structured medical data while preventing unintended or unsafe operations.

We have a few questions and would appreciate the community's guidance:

1. Feasibility: Can NeMo Guardrails be effectively used for Text-to-SQL transformations? Are there any examples, best practices, or documentation to help us get started?
2. Pre-trained Models: Does the NVIDIA ecosystem provide pre-trained models or APIs optimized for natural language understanding in this specific context which can be used inside NeMo Guardrails?
3. Domain-Specific Insights: Are there any examples or lessons learned from using NeMo in healthcare, especially for accessing sensitive and critical data like medication databases with HTTP Requests.

We believe that integrating Text-to-SQL with NeMo Guardrails will greatly enhance the efficiency of healthcare professionals by providing quick, secure access to crucial information.