Secure Text-to-SQL Interface

Transforming data access with a natural language interface that converts queries into SQL, simplifying database interaction.

FMCG Retail

2023

Client and Industry

An innovative egg retail start up in India has engaged our services to enhance their data querying capabilities

Industry

FMCG Retail

Business Type

Start-up

Services

Development & Deployment

Our Role

We have developed an LLM-based interface that effectively translates natural language into SQL queries, seamlessly bridging the gap between conversational input and database interactions.

Project Goals

01

Develop a Privacy-Preserving Text2SQL System: Create a system that accepts natural language queries and database schema, and outputs the corresponding SQL query.

02

Automatic Query Execution: Enable the generated SQL queries to be executed automatically on the existing database to retrieve the required results.

03

System Adaptability: Ensure that the system can be adapted for use with various database systems beyond MySQL to guarantee broad applicability.

Business
Challenges

01.

Data Privacy: Data privacy is a paramount concern, as existing tools that offer similar functionalities tend to compromise security by exposing database content to LLMs for generating and executing SQL queries.

02.

Scalability with Data Size: Managing large data sizes effectively, which can become problematic as the database grows.

Solution

We developed a robust system that processes natural language queries and database schemas as input into an LLM, which then generates the corresponding SQL query. This query is executed in the backend to produce the desired result

Key Features

01

Privacy Protection: The LLM does not access actual data, ensuring stringent privacy.

02

Flexible Data Hosting: Data can reside on either our servers or the client’s servers. Our system functions without direct data access; only the database schema is required, addressing the challenge of handling large data sizes.

03

Adaptable to Various Databases: While our current solution utilizes a MySQL database, our methodology is adaptable to other systems such as ElasticSearch, enhancing versatility.

Impact

01.

Our client successfully implemented the tool to foster an organization-wide data-centric culture for decision-making. This empowerment enables every team member to query data directly and ask meaningful questions, significantly enhancing their operational efficiency and strategic planning capabilities.