Documentation
Three guides to get you from install to productive in under 10 minutes.
Getting Started
Install RowLeap, load your first file, and run your first query. This guide walks through everything from download to your first result set — it takes about 5 minutes.
What you'll learn
- ✓ Installing RowLeap on your platform
- ✓ Loading CSV, SQLite, and Parquet files
- ✓ Running your first SQL query
- ✓ Exporting results
1. Download and install
Go to the Download page and grab the installer for your platform. RowLeap is a native app — no browser required, no Node.js, no Python. Double-click the installer and you're done.
2. Load your first file
Drag any CSV, SQLite, or Parquet file onto the app window — or use the sidebar's "Open file" button. RowLeap reads the file into an in-memory DuckDB database. For CSV files, it auto-detects column types.
3. Run your first query
The file becomes a table named after the file (e.g., sales_data.csv → sales_data). Type SQL into the editor and press Cmd+Enter (macOS) or Ctrl+Enter (Windows/Linux) to run it.
4. Export results
Click the Export button in the results panel and choose CSV, JSON, Parquet, or Markdown. RowLeap writes the file wherever you specify — no cloud upload, no copy-paste.
Querying CSV, SQLite & Parquet Files
RowLeap uses DuckDB under the hood, which means you get a real SQL dialect with full support for JOINs, aggregations, CTEs, and window functions.
What you'll learn
- ✓ How RowLeap names tables from file names
- ✓ Loading multiple files and joining them
- ✓ Querying Parquet and SQLite alongside CSV
- ✓ Handling nulls, type coercion, and date parsing
- ✓ Saving and reusing queries
Table names
RowLeap strips the file extension and replaces spaces, hyphens, and dots with underscores. my-data.csv → my_data. You can always check the table name in the sidebar.
Multiple files
Load as many files as you like — each becomes a separate table. JOIN them just like you would in a real database: SELECT * FROM orders JOIN customers ON orders.customer_id = customers.id.
Parquet and SQLite
Parquet files are loaded the same way as CSV. For SQLite databases, RowLeap reads all tables from the database file and registers each one separately. Query them by table name.
DuckDB SQL dialect
You have access to DuckDB's full SQL dialect: CTEs, window functions, UNNEST, list operations, STRUCT types, regex functions, and more. See the DuckDB SQL docs for reference.
Natural Language → SQL
Type what you want to know in plain English. RowLeap uses built-in AI to generate a SQL query, shows it to you before running, and lets you edit or save it.
What you'll learn
- ✓ How NL→SQL works in RowLeap
- ✓ Writing prompts that produce accurate SQL
- ✓ Reviewing and editing generated queries
- ✓ Iterating with follow-up prompts
- ✓ When to use NL→SQL vs. writing SQL directly
Enable NL mode
NL→SQL is built in — no API key or setup required. Just toggle NL mode on and start asking questions.
Write a prompt
Switch to the NL tab in the query panel. Type a plain English question: "Show me the top 10 customers by revenue last quarter" or "Count rows where status is null". RowLeap sends your schema context + prompt to the AI.
Review before running
The generated SQL appears in the editor before any query is executed. Read it. Edit it if needed. Then run it. You always stay in control of what executes.
Tips for better results
Be specific: mention column names you know, expected output shape, and any filters. "Customers in region = 'West' ordered by lifetime value descending" produces better SQL than "top customers".
Still have questions?
Open a GitHub issue or email us directly.