QueryBot is a modern Generative AI SQL chatbot that empowers non-technical users to query structured databases using natural language — via text or voice input. Built for simplicity, it removes the need to write SQL, making data accessible to everyone.
- Designed for non-technical business users (e.g., managers, analysts)
- Uses a pre-loaded database securely connected via Azure SQL
- Only read operations are allowed — no data mutation
- Users can ask queries using text input or microphone
- Schema, KPIs, and domain prompts are pre-configured
- Lets users pick a data domain to query
- Loads schema, KPIs, and domain-specific prompt context
Examples:
- Sales Performance
- Customer Support Tickets
- Employee Productivity
- Translates fuzzy natural language like:
- “Compare sales this quarter vs last.”
- “What’s the average salary in Marketing?”
- Injects schema/KPI context into GPT prompts for accuracy
- Converts SQL output into business-friendly summaries
- Adds interpretation (e.g., vs targets or previous periods)
Example:
“In March, North region achieved $500K in sales, falling short by $100K from the target of $600K.”
- Users can type or speak their query using a mic
- Voice input is transcribed using Azure Speech-to-Text
- Ideal for accessibility and hands-free scenarios
| Layer | Technology/Library | Purpose |
|---|---|---|
| Frontend | React | UI framework |
| Material UI (MUI) | Components, theming, icons | |
| React Router | Routing/navigation | |
| Chart.js, Recharts | Visualizations | |
| Backend | FastAPI | REST API framework |
| SQLAlchemy | ORM + Azure SQL integration | |
| Pandas | Data transformation | |
| Azure OpenAI (GPT-4) | SQL generation + summarization | |
| Azure Speech-to-Text | Voice input transcription | |
| DevOps | Docker, dotenv | Containerization, config management |
| Security | CORS, Azure Identity | API access + environment security |
- Create a
.envfile with your Azure SQL, OpenAI, and Speech Service credentials. Example:DATABASE_URL="mssql+pyodbc://user:password@server.database.windows.net/database?driver=ODBC+Driver+18+for+SQL+Server" AZURE_OPENAI_API_KEY="your_azure_openai_api_key" AZURE_OPENAI_ENDPOINT="https://your-openai-resource.openai.azure.com/" AZURE_SPEECH_KEY="your_azure_speech_key" AZURE_SPEECH_REGION="your_azure_speech_region" - Install Python dependencies:
pip install -r backend/requirements.txt
- Start the backend:
uvicorn backend.main:app --reload
- Navigate to the frontend directory:
cd frontend - Install dependencies:
npm install
- Start the frontend:
npm start
-
Open your browser and go to: http://localhost:3000
-
Select a data domain from the
Dashboard Explorer -
Go to the
Chattab -
Type your question or click 🎙️ to speak it
-
View:
- The generated SQL query
- Tabular results
- Natural language summary
-
Download results as
xlsx
| Natural Language Input | SQL Translation |
|---|---|
| List all departments | SELECT DISTINCT department FROM employees; |
| Top 5 products by sales in 2023 | SELECT TOP 5 product_name, SUM(sales) FROM orders WHERE YEAR(order_date) = 2023 GROUP BY product_name ORDER BY SUM(sales) DESC; |
| Average salary by department | SELECT department, AVG(salary) FROM employees GROUP BY department; |
| Show total revenue per year | SELECT YEAR(order_date), SUM(revenue) FROM sales GROUP BY YEAR(order_date); |
-
Be specific: Use filters and known fields
-
Use keywords like “compare”, “trend”, “highest”
-
Avoid vague queries like “Show me insights”
-
Only supports read operations — no writes/updates
-
May struggle with vague or ambiguous input
-
Sensitive data masking is not implemented
-
Schema changes require backend reload
See REFLECTIONS.txt for detailed insights.
GPT-4 performs reliably when grounded with schema + KPI context. Voice input makes the tool more inclusive, while summaries make it understandable for business users.
- Azure OpenAI Service
- Azure SQL Database
- Azure Speech Services
- React & Material UI community