DataNova

🌐 Live Demo: https://datanova.streamlit.app

AI-powered autonomous business intelligence platform. Upload data, ask questions in natural language, and get instant dashboards, insights, and executive summaries.

---

Why DataNova

Traditional analytics tools require technical expertise to extract insights from data. DataNova makes business intelligence conversational, autonomous, and accessible through AI-driven analytics workflows.

Architecture

User Query
   │
   ▼
┌───────────────────┐
│  LangGraph Agent  │  ← Intent classification, SQL/Python routing
│  Orchestrator     │
└───────┬───────────┘
   │         │
   ▼         ▼
┌────────┐ ┌──────────┐
│  SQL   │ │  Python  │  ← AST-validated sandbox
│ Engine │ │  Sandbox │
└───┬────┘ └────┬─────┘
   │           │
   └─────┬─────┘
         │
         ▼
┌───────────────────┐
│  Analytics Engine │  ← Plotly, KPIs, data quality
│                   │
│  Dashboard Gen.   │  ← Auto-generated zone grid
│                   │
│  AI Narration     │  ← Executive summaries, trends
└───────────────────┘

Key Features

• ✅ Natural Language to SQL — Ask questions in plain English
• ✅ Autonomous Dashboard Generation — KPIs, charts, data quality, filters
• ✅ AI Executive Summaries — Narrative insights from your data
• ✅ Dual Execution Engine — SQL for queries, Python sandbox for ML/forecasting
• ✅ Dynamic KPI Detection — Automatic metric identification
• ✅ Conversational Analytics — Multi-turn chat with memory
• ✅ Multi-Agent Workflow — LangGraph-powered orchestration

Stack

Layer	Technology
Frontend	Streamlit, Plotly
AI/Orchestration	LangGraph, Groq (LLaMA)
Backend	SQLAlchemy, Pandas, NumPy
Security	AST-based Python sandbox, SQL injection prevention
Deployment	Docker, Streamlit Cloud

Quick Start

git clone https://github.com/Aryaajaiswal/DataNova.git
cd DataNova
pip install -r requirements.txt
echo "GROQ_API_KEY=your_api_key_here" > .env
streamlit run app.py

Docker

docker compose up -d

Challenges & Learnings

• Balancing autonomous dashboard generation with visual clarity
• Improving reliability of LLM-generated SQL through self-correction loops
• Designing schema-aware AI workflows for multi-table databases
• Implementing secure Python execution via AST validation
• Optimizing chart selection dynamically based on data cardinality

Roadmap

• Conversational dashboard editing ("add forecasting", "compare weekdays")
• Real-time database integrations (PostgreSQL, MySQL)
• Multi-user collaboration with saved dashboards
• AI anomaly detection and alerting
• Forecasting engine with ARIMA/Prophet

License

MIT