Tiozin

ETL shouldn't require 80 files, 50 YAMLs, and a PhD in complexity.

Tiozin brings it back to basics: Transform. Input. Output. Nothing more, nothing less.

A lightweight Python framework that makes data jobs declarative, testable, and actually enjoyable to write.

Who is Tiozin for?

Tiozin is human-readable and machine-generatable:

• Data engineers who want reusable pipeline components
• Teams that value declarative jobs
• Projects that require testable ETL logic
• Pipelines where metadata is connected to the execution model

One-Minute Example

pip install tiozin

Using Python directly

from tiozin import TiozinApp

app = TiozinApp()
app.run("examples/jobs/dummy.yaml")

Done. No ceremony, no boilerplate.

Or define a declarative job

kind: LinearJob
name: example_job
owner: tiozin@tiozin.com
maintainer: tiozin
cost_center: tio_scrooge

org: tiozin
region: latam
domain: marketing
subdomain: campaigns
layer: refined
product: users
model: customers

runner:
  kind: NoOpRunner
  name: postgres
  log_level: "{{ ENV.LOG_LEVEL }}"
  db_url: postgres://user:{{ SECRET.FAKE_PASSWORD }}@host:5432/dbname

inputs:
  - kind: NoOpInput
    name: load_it
    layer: raw
    path: .output/lake-{{domain}}-{{layer}}/{{product}}/{{model}}/date={{ DAY[-1] }}

transforms:
  - kind: NoOpTransform
    name: process_it
    strategy: sha256

outputs:
  - kind: NoOpOutput
    name: save_it
    path: .output/lake-{{domain}}-{{layer}}/{{product}}/{{model}}/{{ today }}

Run it:

$ tiozin run examples/jobs/dummy.yaml

Philosophy

Your uncle's advice: keep it simple, readable, and testable.

Tiozin is built around a small set of principles that are not features but constraints that shape the design.

• Declarative: Define what, not how. The YAML format is intentionally minimal, making it easy for both humans and LLMs to generate and maintain jobs.
• Pluggable: Swap runners, registries, and plugins when needed
• Metadata-native: Execution and metadata walk together
• Observable: Logs that actually help
• Testable: Mock anything, validate everything

Documentation

Writing Declarative Jobs

• Quick Start
• Working with Jobs
• Examples

Concepts

• Family Model: Tios and Tiozins
• Jobs
• Batches
• Runners
• Inputs, Transforms & Outputs
• Registries

Extending Tiozin

• Creating a Provider Family
• Creating Pluggable Tiozins
• Tio Proxy: Adding Cross Cutting Family Features
• Tio Kernel - The Core Family
• Tio Spark - Spark Family Example
• Tio Duckdb - DuckDB Family Example

How-to Guides

• How to Work with Incremental Backlogs
• How to Configure OpenLineage
• How to Use Schemas in Jobs
• How to Use Secrets in Jobs

Reference

• Settings Guide
• Settings: YAML Reference
• Settings: Environment Variable Reference
• Templates Reference
• API Reference
• Dependencies

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

License

This project is licensed under the Mozilla Public License 2.0.