DataSense AI - Project Showcase

What is DataSense AI?

DataSense AI is a desktop application that lets users analyze and interact with their data using plain English. Whether your data lives in a SQL Server database or a CSV file, you can just ask questions in natural language and get instant answers—complete with tables, charts, and even the Python code that was generated to get those results.

Key Features

• Natural Language Interface: No need to write SQL or Python. Just type your question and get answers.
• Flexible Data Connections: Connect to SQL Server or select CSV files with a few clicks.
• Smart Code Generation: Uses advanced AI (Mistral AI) to write and generate Python code for your questions.
• Instant Visuals: Automatic charts and graphs help you understand your data at a glance.
• Modern UI: Built with PySide6 and QFluentWidgets for a smooth, responsive experience.
• Real-Time Results: Handles queries in real time, streaming results as soon as they’re ready.
• Extensible Design: Modular architecture makes it easy to add new features or data sources.

How Does It Work?

1. Data Layer

• Connects to SQL Server (via PyODBC) or loads CSV files (via pandas).
• Fetches schema and sample data for analysis.

2. AI Layer

• DataSense-TS2505: Table Selection Agent integrates with Mistral AI to select most relevant table according to query.
• DataSense-MD2505: Datasense AI Agent integrates with Mistral AI to generate Python code for user queries.

3. Processing Layer

• Query Handling: Interprets user questions, generates and executes Python code, and processes results.
• Search: Finds relevant tables/columns using embeddings for smarter query understanding.

4. Presentation Layer

• User Interface: Built with PySide6 and QFluentWidgets for a modern look and feel.
• Results Display: Shows tables, charts, and generated code in real time.
• Error Handling: User-friendly messages for connection issues, invalid queries, or code errors.

Typical User Flow

1. Connect Your Data:
   Choose SQL Server or upload a CSV file.

2. Ask Questions:
   Type your data questions in plain English.

3. See Results:
   Instantly view tables and charts generated on the fly.

4. Tweak Settings:
   Adjust appearance and preferences in the settings menu.

Screenshots

Splash Screen:

Data Source (MSSQL) :

Data Source (CSV) :

Results Page:

Results Page (Plot) :

Light Theme:

Technical Stack

• PySide6: Desktop GUI
• QFluentWidgets: Modern UI components
• pandas: Data wrangling
• pyodbc: SQL Server connectivity
• matplotlib: Charting and visualization
• sentence-transformers: Semantic search and embeddings
• Mistral AI: AI-powered code generation

API Switch Configuration (api_switch.json)

The api_switch.json file lets you set a custom API key for MistralAI, which is the only supported backend right now. The app comes with a default MistralAI API key built in, but if you want to use your own key (for privacy or higher limits), just put it in this file. The app will check for your custom key at startup and use it if found.

• Location: Root directory of the project
• Format: Simple JSON with a key "mistralKey" (e.g., { "mistralKey": "your-mistral-key-here" })
• Usage: If you add your own API key here, the app will use it for all MistralAI requests. If not, it falls back to the default key.
• Why: Makes it easy for users to use their own MistralAI account, and helps with privacy or API limits.

Project Structure

src/
├── core/
│   ├── data_source.py          # Data source integration (SQL, CSV)
│   ├── ai_agents.py            # AI agents for code generation
│   ├── embedding_generator.py  # Schema and data embeddings
│   ├── query_processor.py      # Query processing and execution
│   ├── response_generator.py   # Formatting and streaming results
│   └── llm_integration.py      # Mistral AI integration
├── gui/
│   ├── main_window.py          # Main application window
│   ├── pages/
│   │   ├── base_page.py        # Shared page logic
│   │   ├── data_source_page.py # Data source setup
│   │   ├── results_page.py     # Results display
│   │   └── settings_page.py    # App settings
├── assets/                     # Images and resources
├── assets.py                   # Compiled assets
├── assets.qrc                  # Resource file
└── main.py                     # App entry point

Error Handling

• Connection and authentication errors are shown in the app.
• Invalid queries or code errors are caught and explained to the user.
• Visualization problems are handled gracefully with clear messages.

Extending DataSense AI

• Add new data sources: Update core/data_source.py
• Integrate different AI models: See core/llm_integration.py
• Change the look and feel: Tweak files in the gui/ folder

Future Improvements

• More data sources (databases, file types, cloud)
• Smarter code generation and query understanding
• Enhanced visualizations and customization
• Better error messages and user feedback

Conclusion

DataSense AI brings the power of natural language and AI to data analysis, making it accessible for everyone—no coding required. Its modular design and robust error handling ensure it’s ready for both end-users and future development.

Prepared for project showcase