Export the pandas/Polars code you'd write by hand — no lock-in.
Build data and ML pipelines on a canvas, preview every step, then export readable pandas, Polars, or lazy Polars code with no proprietary runtime. What you keep is real, reviewable Python you can run and version anywhere.
Docs · Quick Start · Plugins · Discussions · Contributing
If this looks like something you'd use, a ⭐ on GitHub genuinely helps a small open-source project like this get discovered.
Ciaren is an open-core, plugin-first platform that exports the pandas or Polars code you'd write by hand for data engineering and lightweight machine-learning workflows — no proprietary runtime, no lock-in.
The canvas is there for speed: build on it, preview every step, then keep the same clean, ordinary Python whether you run it inside Ciaren or take it somewhere else.
python -m pip install ciaren
ciaren serveOpen http://localhost:8055. The PyPI wheel bundles the React editor, so users
do not install a separate frontend package. One install starts the API,
scheduler, and web UI.
Alpha software. Ciaren is under active development. APIs, workflow formats, generated code, plugin interfaces, and internal data models may change before
1.0.0. Use it for learning, prototypes, and controlled internal workflows before relying on it for critical production jobs.
| You get | Why it matters |
|---|---|
| Visual workflow building | Design pipelines faster than writing every dataframe step by hand. |
| Live node previews | Inspect data samples and schema changes before running the full flow. |
| Clean Python export | Generate standalone pandas, Polars, or lazy Polars scripts you can review and run outside Ciaren. |
| Local-first execution | SQLite works out of the box, and your data does not need to leave your machine. |
| Data engineering + ML | Ingest, clean, validate, engineer features, train, evaluate, predict, and export from one canvas. |
| Plugin-first architecture | Add custom nodes, connectors, engines, model providers, validators, and exporters outside core. |
Ciaren is not a hosted no-code black box. Every node maps to understandable dataframe behavior, every run leaves inspectable results, and every flow can become code.
python -m pip install --upgrade pip
python -m pip install ciaren
ciaren serveOpen http://localhost:8055, then open Projects -> Demo. The first start
seeds sample datasets and working example flows, so you can preview, run, and
export something real before uploading your own data.
For repeatable evaluation, pin the release version:
python -m pip install "ciaren==0.2.0"git clone https://github.com/ciaren-labs/Ciaren.git
cd Ciaren
docker compose up --buildOpen http://localhost:8055.
Use this path when contributing to the backend, frontend, docs, or plugin SDK.
Requirements: Python 3.12+, Node.js 18+, and Git.
git clone https://github.com/ciaren-labs/Ciaren.git
cd Ciaren/backend
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e .
ciaren serveIn a second terminal:
cd Ciaren/frontend
npm install
npm run devOpen http://localhost:5173 for the live development frontend. The backend API
and Swagger docs stay on http://localhost:8055.
- Input and output: CSV, TSV, Excel, Parquet, JSON/JSONL, text, SQL databases, S3, GCS, and Azure Blob.
- Cleaning: drop/fill nulls, remove duplicates, rename/select/drop columns, cast types, replace values.
- Transformation: filters, joins, group by, aggregate, calculated columns, maps, pivots, windows, sorting, sampling.
- Data quality: assert not-null, unique, value ranges, row count, expressions, and allowed values.
- Machine learning: split, train, cross-validate, predict, evaluate, feature engineering, importance, and MLflow tracking.
- Operations: run history, scheduling, REST API, CLI, webhook trigger, and Python SDK.
A simple read -> clean -> aggregate -> write flow can export to code like this:
import polars as pl
df_1 = pl.read_csv("sales.csv")
df_1 = df_1.drop_nulls(subset=["amount"])
df_1 = df_1.group_by(["region"]).agg([pl.col("amount").sum().alias("amount")])
df_1.write_csv("summary.csv")The same flow can also export pandas or lazy Polars variants where supported. That makes Ciaren useful for prototyping, teaching, code review, and migration from visual exploration into normal Python projects.
Ciaren's open core stays focused on the shared workflow platform. Specialized integrations should usually be plugins, not one-off patches to core.
| Extension point | Examples |
|---|---|
| Nodes | Custom transforms, validators, AI-assisted steps, domain-specific operations |
| Connectors and storage | Internal APIs, SaaS tools, warehouses, object stores, document databases |
| Model providers | Local models, scikit-learn estimators, organization-specific training logic |
| Execution engines | Alternative dataframe engines and future runtime targets |
| Exporters | Code targets, deployment bundles, validation reports |
Plugins can be packaged as .ciarenplugin files, signed, inspected, installed,
enabled, disabled, and distributed independently.
Start here:
- Data analysts: clean, reshape, validate, and export datasets without writing every step by hand.
- Data engineers: prototype repeatable local pipelines, inspect generated code, then automate with CLI/API/SDK workflows.
- Python learners: see how visual dataframe operations become pandas and Polars code.
- ML practitioners: try lightweight ML flows with local MLflow tracking.
- Plugin authors: ship custom nodes, connectors, engines, and model providers without maintaining a fork.
- Open-source contributors: help polish the editor, execution engine, transformations, docs, tests, and plugin SDK.
- Installation - PyPI, Docker, source installs, extras, and troubleshooting.
- Quick Start - build your first flow in five minutes.
- Demo Project & Tutorials - walk through the seeded example flows.
- Examples - end-to-end workflow walkthroughs.
- Machine Learning Quick Start - train and evaluate a model on the canvas.
- Plugin Guide - build your first plugin.
- Roadmap
- Security
- Contributing
The base install includes the core app, pandas, Polars, scikit-learn models, MLflow tracking, and the bundled web UI.
Install extras only when you need specific drivers or optional model families:
python -m pip install "ciaren[postgres]"
python -m pip install "ciaren[s3]"
python -m pip install "ciaren[ml]" # XGBoost and LightGBM choices
python -m pip install "ciaren[keyring]" # recommended: store connection secrets in the OS keychain
python -m pip install "ciaren[signing]" # plugin signing supportOn a desktop install, ciaren[keyring] is the recommended way to hold
database and API secrets: the connection form (and the ciaren secret set
command) stores the value in your OS keychain — Windows Credential Manager,
macOS Keychain, or Secret Service on Linux — and the connection keeps only a
keyring:NAME reference. Ciaren never writes the secret to its database.
Without the extra, connections still work with environment-variable
(env:/PG_PASSWORD) or secret-file (file:/run/secrets/...) references.
The separate ciaren-client package is only for scripts that need to call a
running Ciaren server. It is not required to use the visual app.
Ciaren is early, and useful contributions are welcome: reproducible bugs, example flows, docs improvements, transformation nodes, plugin SDK improvements, frontend workflow polish, tests, and focused core fixes.
The open core is intentionally lightweight. New niche databases, SaaS products, internal APIs, proprietary storage systems, and organization-specific model logic should normally be built as plugins. If the SDK blocks that work, open an SDK-focused issue or discussion.
Start with CONTRIBUTING.md. Questions and ideas can go to GitHub Discussions, and reproducible bugs or focused feature requests can go to GitHub Issues.
Review exported Python code, test flows before using important data, and add appropriate operational controls before using Ciaren in sensitive or critical environments.
Please report vulnerabilities using the process in SECURITY.md.
- Ciaren Core: AGPL-3.0-only.
- Public Plugin API / SDK: Apache-2.0.
- Plugins: may use their own compatible license, depending on the plugin author and distribution model.
- Future cloud or hosted services: not necessarily covered by this open source repository license.
In plain language: using Ciaren — locally or self-hosted inside your organization — does not obligate you to open-source anything, your flows and exported Python are yours, and the Apache-2.0 SDK means plugin authors choose their own license. And a standing commitment: capabilities that ship in the open core stay in the open core — commercial offerings, if they ever exist, add on top rather than carving features out.
See LICENSE, NOTICE, and LICENSES/ for the complete license texts and notices, and the licensing FAQ for common questions.
If Ciaren is useful in your research, teaching, or writing, please cite it using the metadata in CITATION.cff (GitHub's "Cite this repository" button on the repo sidebar generates APA/BibTeX from the same file):
@software{ciaren,
author = {Arenas, Rodrigo},
title = {Ciaren},
url = {https://github.com/ciaren-labs/Ciaren},
license = {AGPL-3.0-only},
version = {0.2.0},
date = {2026-07-20}
}- Current stage: Alpha (pre-1.0) — the API and data model may change
before
1.0.0. Pin a version for production use. - Latest release:
0.2.0(first public release was0.1.0) — see the CHANGELOG for what's included and for breaking changes. - Breaking changes are expected before
1.0.0
Made for data practitioners who value local workflows, transparent execution, and Python they can actually read.