Ciaren

v0.3.0 alpha · open source, AGPL-3.0

Open-source visual ETL that exports pandas and Polars code

Build data pipelines on a canvas that runs on your machine, check each step on real data, and export a Python script you can read and run without Ciaren.

Get startedGitHub

Quick start

Runs on your machine

  1. Install from PyPI

    python -m pip install ciaren

  2. Start the editor

    ciaren serve

  3. Open it in your browser

    http://localhost:8055

Installation guide
Flow editor
Building an order-status prediction flow in the editor, with a data preview at each step.

Code export

The canvas and the code it exports

Each node on the canvas becomes a few lines of pandas or Polars. The exported script runs without Ciaren installed.

  1. File Input

    sales.csv

  2. Filter Rows

    status == 'completed'

  3. Group By Aggregate

    sum(amount) by region

  4. Rename Columns

    amount -> total_sales

  5. Sort Rows

    total_sales, descending

  6. File Output

    sales_by_region.csv

pipeline.py
import polars as pl df_sales = pl.read_csv('sales.csv') df_sales = (    df_sales.filter(pl.col('status') == 'completed')    .group_by('region')    .agg(pl.col('amount').sum())    .rename({'amount': 'total_sales'})    .sort('total_sales', descending=True, nulls_last=True)) df_sales.write_csv('sales_by_region.csv')

The same flow exports to either engine. You pick pandas or Polars per run.

What it does

From the canvas to a script you can run anywhere Python runs

Preview

Check every step on your real data

Select any node to see its output before you run the rest of the flow. A bad join or a wrong filter shows up while you build, not after a full run.

  • Rows, columns, and types for the selected node
  • Dataset profiles with types, null counts, and distributions
Tour the editor
Flow editor · data preview
Ciaren editor with a canvas of connected nodes, the node palette, the configuration panel, and a live data preview table

Export

Take the flow with you as Python

Export the flow as a readable script. Run it in a notebook, a cron job, or CI.

  • pandas, Polars, or lazy Polars output
  • A standalone script that runs without Ciaren installed
How code export works
Export code
Screen recording of the Generated code dialog switching between pandas, Polars, and lazy Polars as the code regenerates from the same flow

Runs and schedules

Know what each run did, and run it again on a schedule

Every run records per-node status and row counts, so a failure points to the step that caused it. Schedules run flows on your own machine.

  • Hourly, daily, weekly, or a custom cron expression
  • Retries, catch-up runs, and auto-disable on repeated failure
Scheduling guide
Runs
Ciaren runs list: every execution with its status badge, engine, trigger, and timestamps

Also in the core

Everything else ships in the open-source install

80 built-in nodes

File, SQL, transformation, chart, and output nodes you drag onto the canvas and connect.

Choose the engine per run

Run the same flow on Polars or pandas without changing the flow.

Connect your data

SQL databases, object storage, files, and REST APIs, with credentials kept explicit and local.

Train and track models

Split, train, predict, and evaluate, with every training run tracked in MLflow.

Check data quality

Assert not-null, unique, range, expression, and row-count checks inside the flow.

Extend with plugins

Add nodes, connectors, and ML model types through the Apache-2.0 plugin SDK.

Alternatives

How it compares

Ciaren sits between notebooks, orchestrators such as Airflow and dbt, and closed visual ETL tools. The comparison guide covers where each one fits, and when not to use Ciaren.

Compare Ciaren with other ETL tools

Project status: pre-1.0 alpha

Ciaren is under active development. APIs, the workflow format, generated code, and plugin interfaces may change before 1.0. It suits experiments, prototypes, and controlled internal workflows.

Try it on your own data

Ciaren is free and open source. Install it with pip and open the editor in your browser.

Get startedGitHub