Lumora

Lumora is a story-first data understanding platform that transforms spreadsheet data into accurate insights and visual narratives. It empowers non-technical users to explore complex datasets using plain English, delivering real-time summaries, anomaly detection, and dynamic charts — all without writing a single line of SQL.

Technologies Used:

• Frontend: Next.js 14, TypeScript, Tailwind CSS, Recharts, Zustand
• Backend: FastAPI (Python), DuckDB, Pandas
• AI: Groq LLM for natural language to SQL translation and data narration
• Validation & Storage: Pydantic, R2 (via Boto3), environment-based secret handling
• Deployment: Vercel (frontend), Render (backend)

Key Features:

• Natural Language Query Interface: Users can ask questions like "Show me sales by region" and receive SQL-backed answers with zero hallucination risk, thanks to Groq’s high-precision language model.
• Sub-Second Query Performance: DuckDB enables real-time SQL execution on datasets with 500K+ rows, ensuring instant results even with large files.
• Automated Insight Engine: Lumora detects trends, spikes, and outliers without user input — surfacing insights like “Sales dropped 38% in March” or “Top 10% of users drive 72% of traffic.”
• Dynamic Data Visualizations: Automatically selects from 8+ chart types based on the data — including bar, line, donut, and area charts — rendered using Recharts in a modern, responsive UI.
• Smart Ambiguity Resolution: If a user asks for “profit” but the dataset uses “net_income,” Lumora detects the mismatch and prompts a correction through an intuitive mapping UI.
• Ephemeral Data Sessions: No long-term storage. Files are processed in-memory and discarded after the session to ensure privacy and security.

Performance & Impact:

• Achieved sub-1 second query latency on 500K+ row datasets using DuckDB
• Resolved 90%+ of user queries through accurate natural language mapping
• Reduced manual data analysis time by up to 70% via automated insights and charts
• Rendered 8+ chart types without any user configuration, enhancing usability

Live Demo:

lumora.pranavzagade.com