Overview

SenseQL is an AI-powered query agent that lets you talk to your databases in natural language.
Instead of writing SQL, just ask questions like:

“Show me all delivered orders in July.”

SenseQL will generate, run, and return SQL results automatically.

Supports:

• PostgreSQL

• Google BigQuery

1. Tech Stack

SenseQL is built with a modern, modular stack:

Frontend (React)

• Login and authentication

• Connect user databases

• Chat interface for asking queries

Backend (NestJS + PostgreSQL)

• User authentication and session management

• Stores connected databases and chat history

• Forwards queries to the GenAI server

GenAI Server (FastAPI + LangChain + LLMs)

• Fetches schema from the user database

• Selects tables and columns

• Generates SQL queries

• Executes SQL against PostgreSQL, BigQuery, or MySQL

2. System Workflow

How SenseQL turns your question into SQL:

1. User Input
   You type a question in the chat.
   → Sent to backend via WebSocket.
   
   

2. Table Selection Agent
   Picks the relevant tables from your schema.
   → Handles ambiguities and may ask clarifying questions.
   
   

3. Column Selection Agent
   Matches query intent with column names.
   → Detects synonyms and typos.
   
   

4. SQL Generation Agent
   Uses an LLM to build a safe, optimized SQL query.
   → Includes retries and error correction.
   
   

5. Query Execution
   SQL runs on your database.
   → Results returned to the frontend and stored in chat.

3. Agents Explained

Table Selection Agent

• Input: natural language + schema

• Output: list of relevant tables

Column Selection Agent

• Input: selected tables + user query

• Output: relevant columns and data types

SQL Generation Agent

• Input: tables + columns + query intent

• Output: final SQL string (with previews and retries)

4. Integrations

SenseQL connects directly to:

• PostgreSQL

• Google BigQuery

Planned: Snowflake, MySQL