Real apps get data from outside; ours needs a database to exist first. A seed script is a plain Crow program — chapter 11, nothing more:

use db

conn = connect['sqlite://trades.db']
exec[conn, 'DROP TABLE IF EXISTS trades']
exec[conn, 'CREATE TABLE trades (id INTEGER PRIMARY KEY, symbol TEXT, qty INTEGER, price REAL, pnl REAL)']
exec[conn, 'INSERT INTO trades VALUES (1, ?, ?, ?, ?)', ['AAPL', 100, 191.5, 250.0]]
exec[conn, 'INSERT INTO trades VALUES (2, ?, ?, ?, ?)', ['MSFT', 50, 415.2, -120.5]]
exec[conn, 'INSERT INTO trades VALUES (3, ?, ?, ?, ?)', ['NVDA', 20, 875.0, 890.2]]
exec[conn, 'INSERT INTO trades VALUES (4, ?, ?, ?, ?)', ['AAPL', 40, 193.1, 96.4]]

counts = query[conn, 'SELECT count(*) AS n FROM trades']
log['seeded `counts.n[0]` trades']
rm['trades.db']

expected output — from the book, verified by crow test --book

seeded 4 trades

(The final line removes the file only so this book's test runner stays tidy — the real seed script, examples/dashboard_seed.co, keeps it.)

sqlite://trades.db is a file this time, not :memory: — the data survives between runs, and the dashboard process will open the same file. Swap in a postgres://... URL later and nothing else changes.

Loading and Shaping the Data