A tiny, typed, sequential pipeline engine for Python.

lilpipe is a lightweight, Pydantic-powered library for building and running sequential workflows in Python. It’s ideal for data processing, scientific workflows, and any task that benefits from a clear, linear sequence of steps. With built-in caching and simple control signals, lilpipe is small, typed, and practical.

• Sequential workflows — run steps in a fixed order.
• Type-safe context — Pydantic-based PipelineContext.
• Smart caching — fingerprint_keys to skip unchanged work.
• Flow control — ctx.abort_pass() and ctx.abort_pipeline().
• Composable steps — nest via Step(name, children=[...]).
• Tiny surface area — minimal API and dependencies.

lilpipe is intentionally not a DAG/orchestrator. It’s a tiny, in-process, sequential runner for code you’d otherwise write as a script.

Use lilpipe when…

• The whole workflow runs top-to-bottom in one process and finishes in seconds/minutes.
• You don’t want extra infra: no server, no scheduler UI, no metadata DB.
• You want a typed, validated shared state (Pydantic PipelineContext) flowing step-to-step with helpful errors.
• You want incremental reruns: fingerprint caching to skip unchanged steps across runs.
• You need simple retry/abort semantics without standing up an orchestrator.
• You like small, testable units: steps you can unit-test and compose (Step("name", children=[...])).

Examples:

• Data science: CSV → clean → feature engineer → fit/evaluate → export plots.
• Lab/assay: load plate → calibrate → QC/validate → generate PDF report.
• Data eng script: pull S3 JSON → normalize → write Parquet → send Slack notification.
• Dev tooling: codegen → format → validate → publish artifact.

pip install lilpipe

See pyproject.toml for supported Python versions.

Run the full demo notebook live on Binder:

(If you’d rather just view it, the raw notebook is at examples/example.ipynb in the repo.)

Licensed under the Apache 2.0 License.

• Issues: https://github.com/andrewruba/lilpipe/issues
• PyPI: https://pypi.org/project/lilpipe/