Skip to content
ETLantic 0.50 · Published Beta

One typed pipeline model. Many execution backends.

Define contracts and topology in Python, validate them before execution, then produce deterministic plans for local engines, backend plugins, or external orchestrators.

Python 3.11+ MIT licensed Typed
ETLantic gives Python data pipelines one portable, typed logical model. Validate-before-write is the safety gate; deterministic planning and pluggable execution carry that model into dataframe engines and orchestrators. Cross-engine execution remains explicit.

Choose your path

Status: Available in ETLantic 0.53.0 (Beta release candidate).

Green path: first success

Install from PyPI first: pip install etlantic. The commands below pin etlantic==0.53.0 so this version of the documentation and the installed API stay aligned. The complete Quickstart continues with an intentional validation failure after the first successful run.

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install 'etlantic==0.53.0'
python -m etlantic --version

mkdir my-pipeline
cd my-pipeline
python -m etlantic init --with-toml
python -m etlantic validate pipeline.py:SamplePipeline --profile development
python -m etlantic run pipeline.py:SamplePipeline --profile development
cat data/out.json
py -3.11 -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
python -m pip install 'etlantic==0.53.0'
python -m etlantic --version

mkdir my-pipeline
cd my-pipeline
python -m etlantic init --with-toml
python -m etlantic validate pipeline.py:SamplePipeline --profile development
python -m etlantic run pipeline.py:SamplePipeline --profile development
Get-Content data\out.json
Expected resultYou should see a succeeded run and two JSON rows for Ada and Grace.

init should run in a fresh directory. --force can overwrite scaffolded files, so use it only when you have reviewed the target directory.

PyPI users and repository examples

The wheel does not include examples/. Pip users should continue to First Pipeline, then choose a PyPI-ready engine tutorial. Contributors can use the repository examples/ after uv sync --locked.

Why validation-first

01 Fail before side effects Catch invalid wiring, incompatible contracts, unavailable capabilities, and plugin-trust failures before publication.
02 Review what will run Inspect, fingerprint, diff, and retain deterministic, secret-free plans as build evidence.
03 Keep execution explicit Use one logical model while plugins own Polars, Pandas, SQL, PySpark, and orchestration behavior.

Choose an execution path

Core installs without dataframe engines, database drivers, Spark, Airflow, or Prefect. Add only what the pipeline uses.

Path Install Current scope
Local Python + JSON/CSV pip install 'etlantic==0.53.0' Built-in first-success and test path
Polars pip install 'etlantic[polars]==0.53.0' Eager/lazy execution and portable compilation
Pandas pip install 'etlantic[pandas]==0.53.0' Eager execution and portable compilation
SQL pip install 'etlantic[sql]==0.53.0' SQLite evaluation and PostgreSQL reference execution
PySpark pip install 'etlantic[pyspark]==0.53.0' Batch Spark execution; compatible JVM required
Airflow pip install 'etlantic[airflow]==0.53.0' DAG compilation; Apache Airflow installs separately
Prefect pip install 'etlantic[prefect]==0.53.0' Bounded local direct-execution integration

See Engine selection for prerequisites and Compatibility before pinning a deployment.

Know the release boundary

ETLantic is Beta, community-supported, with no SLA. Use it for documented single-tenant pilots. You can embed an HTTP control plane (etlantic-fastapi) with Supported isolation profiles (isolated-deployment, dedicated-schema). There is no hosted multi-tenant SaaS.

Available Typed authoring, validation, deterministic planning, local and plugin execution, Airflow compilation, schedules, human-governed AI context and proposals, reports, and observability providers.
Experimental Kafka, Iceberg, MCP, DataFusion, Kubernetes, Spark Connect, and shared-service isolation may change or stay fake-first.
Not included Hosted multi-tenant SaaS, a formal SLA, or compliance certification.

Production profiles require a non-empty plugin_allowlist; allowlisting controls selection, not process isolation. Plans and reports carry secret references, never resolved values. Review Capabilities, Security, Production readiness, and the planned multi-tenant control-plane program before a pilot.

Continue by goal

Goal Next document
Understand the logical model Core concepts
Author with the public Python facade Python SDK in 10 minutes
Configure CI validation CI integration
Diagnose a failure Troubleshooting
Review the current release What's new in 0.53
Review future direction Planning Hub
Contribute to ETLantic Contributor guide