Documentation

Introduction to Visual

Visual is an AI-powered data visualization platform. You describe what you want to see in plain English, and Visual writes SQL, explores your data, and generates charts automatically.

The AI agent understands your data structure, runs exploratory queries to find patterns and anomalies, then produces the right visualization type for your question.

How It Works

When you ask a question, the Visual agent follows a structured workflow:

1. Data exploration — reads your schema, samples rows, checks column types and distributions.
2. Pattern detection — runs aggregations, looks for trends over time, finds correlations.
3. Anomaly detection — flags outliers or data quality issues before charting.
4. Visualization planning — chooses the right chart type based on your data and question.
5. Chart generation — produces one or more charts with clear titles, labels, and key insights.

Quickstart

1. Drag and drop a CSV file into the composer, or type @ to attach a file.
2. Ask a question like "Show me sales by region" or "What's the trend over time?"
3. Visual will analyze your data and generate charts. You can refine by asking follow-up questions.
4. Save your visualization to a project, share it with your team, or export it.

File Uploads

Upload CSV, TSV, or Parquet files. Visual converts them to an internal columnar format (Parquet) for fast SQL queries via DuckDB. Column types are inferred automatically.

Multi-File Analysis

You can upload multiple files to a single visualization. Visual creates separate tables for each file, and you can JOIN them in SQL queries. The AI agent will discover all available data sources and query across them.

SQL Queries

Under the hood, Visual uses SQL to query your data. The agent generates queries like:

SELECT region, SUM(sales) as total_sales
FROM "your_file_abc123"
GROUP BY region
ORDER BY total_sales DESC

You don't need to write SQL yourself — just describe what you want. But if you're comfortable with SQL, you can view and edit the generated queries.

Basic Charts

Visual supports a wide range of chart types:

• Bar — compare values across categories
• Line — show trends over time
• Area — like line charts but with filled regions
• Scatter — show correlation between two numeric variables
• Bubble — scatter with a third dimension encoded as bubble size
• Pie / Doughnut — show proportions of a whole

Advanced Charts

• Heatmap — two categories with intensity
• Treemap — hierarchical data as nested rectangles
• Sankey — flow between nodes
• Radar — multivariate data on radial axes
• Candlestick — financial OHLC data
• Gauge — single KPI with target ranges

Maps

Choropleth maps color geographic regions by value. Upload data with country or region codes, and Visual will render an interactive map.

Teams

Create a team to share visualizations with colleagues. Team members can view, edit, and collaborate on projects. You can organize work into projects within a team.

Sharing

Make a visualization public to share it with anyone via link. Public visualizations are view-only. For private sharing, add collaborators to your team.