LAVA — LLM-Assisted Visual Analytics

Ask natural language questions about your data, get interactive visualizations.

Setup

Requirements: Python 3.13+, UV

# Clone and install
git clone <repo-url>
cd BigData-Visualization
uv sync

First Run

uv run python main.py

On first run you'll be prompted for your OpenAI API key:

No OpenAI API key found.
Get one at: https://platform.openai.com/api-keys

Enter your OpenAI API key: sk-...
Key saved to .env

The key is saved to .env and loaded automatically on future runs.

Then open http://127.0.0.1:8050 and start asking questions.

Usage

uv run python main.py                        # Default: sample sales data
uv run python main.py --data path/to/file.csv # Your own CSV or Parquet
uv run python main.py --port 3000             # Custom port

Example Queries

• "Show total revenue by region"
• "Revenue trends over time, monthly"
• "What's the distribution of unit prices?"
• "Compare revenue per product category as a radar chart"
• "Top 10 products by revenue"

Development

uv run pytest              # Run tests (71 tests)
uv run ruff check .        # Lint
uv run ruff format .       # Format

Architecture

User query → [LLM Refine] → [LLM VizSpec] → [SQL Gen] → [DuckDB] → [Plotly] → Browser

Three layers: Data Engine (DuckDB) | LLM Pipeline (GPT-4o-mini) | Viz Renderer (Plotly/Dash)

Team

CS 43016 — Big Data, Kent State University Joseph Demore, Randy Truong, Nathan Naples, Annika Hall, Cristian Quezada, Kregg Jackson