Start as a script. Ship to your real app.

It starts as the one Python script you'd already write. As it grows, n6k splits the compute out into a cached data.py and lets the view become Markdown or plain React. So unlike Streamlit, Dash, or Panel, you never hit the wall where the app is the render loop, with nowhere to add auth or your own frontend.

uvx n6k init
uv run n6k dev

page.py

start here

Your whole app in one Python file — the script you'd write anyway. It recomputes on every interaction.

@page
def revenue(c: Components) -> None:
    df = pd.read_parquet("orders.parquet")   # re-read every rerun
    by_month = df.groupby("month", as_index=False).revenue.sum()
    c.Static.BarChart(rows=by_month.to_dict("records"), x="month", y="revenue")

As it grows, move the data work into a data.py — now it's fast

data.py

your data layer

Define the data once in Python. n6k caches it and runs the SQL live in the browser.

@db.data()
def orders() -> pd.DataFrame:
    return pd.read_parquet("orders.parquet")

@orders.view()
def by_month() -> str:
    return "SELECT month, sum(revenue) AS revenue FROM orders GROUP BY month"

The same chart over that one data.py — your choice of syntax.

<BarChart data="db.by_month" x="month" y="revenue" />

Grows into a real app

n6k.py

your backend

Add auth — it's plain FastAPI middleware over your own app.

from n6k_app import create_app

app = create_app()

@app.middleware("http")
async def auth(request, call_next):
    # your auth check here
    return await call_next(request)

app/revenue/page.tsx

your frontend

Or drop the same chart straight into your Next.js app.

"use client";
import { QueryProvider, DuckDBProvider } from "@n6k.io/db/react";
import { BarChart } from "@n6k.io/ui/components";

export default function Page() {
  return (
    <QueryProvider>
      <DuckDBProvider databases={{ db: "n6k://your-server/p/revenue" }}>
        <BarChart data="db.by_month" x="month" y="revenue" />
      </DuckDBProvider>
    </QueryProvider>
  );
}