Skip to content

Repository files navigation

lythonic

PyPI version Documentation

Compose Python logic into data-flow pipelines — sync or async, run anywhere.

Write plain Python functions. Wire them with >>. Data flows visibly between nodes — you can see what went in, what came out, what failed. Unlike task schedulers where jobs are opaque units, lythonic tracks the data itself.

Quick Start

uv add lythonic
from lythonic.compose.namespace import Dag

def fetch(url: str) -> dict:
    return {"source": url, "values": [1, 2, 3]}

def double(data: dict) -> dict:
    return {**data, "values": [v * 2 for v in data["values"]]}

dag = Dag()
dag.node(fetch) >> dag.node(double)

# Run it — sync or async, doesn't matter
import asyncio
result = asyncio.run(dag(url="https://example.com"))
print(result.outputs)  # {"double": {"source": "...", "values": [2, 4, 6]}}

Why lythonic?

Data flow, not task flow. Each node receives typed data from upstream and passes results downstream. The DAG runner wires inputs to outputs by type — fan-out, fan-in, and map-reduce built in. Provenance tracking records what data flowed through each edge.

Compose freely. DAGs nest inside DAGs. dag.node(sub_dag) runs a sub-DAG as a single step. dag.map(sub_dag) runs it on each element of a collection, concurrently. Build small, reuse everywhere.

Run transparently. await dag() for a quick test. DagRunner with provenance for production. lyth start for a long-running engine with cron-triggered pipelines. Same code, different execution context.

Sync and async — mixed freely. Write sync functions, async functions, or both in the same DAG. Sync nodes run in a thread executor automatically.

Features

  • DAG composition>> wiring, callable DAGs, MapNode, CallNode
  • @dag_factory — define reusable DAG templates as decorated functions
  • Triggers — cron-scheduled or push-triggered execution via TriggerManager
  • Provenance — SQLite-backed tracking of runs, node executions, edge traversals
  • Caching — per-callable SQLite cache with probabilistic TTL refresh
  • lyth CLIstart, stop, run, fire, status commands
  • State — Pydantic-based SQLite ORM with schema management and multi-tenant support
  • Portable data typesFrameData, ExposureMatrix, SymmetricMatrix, KeyedVector: JSON in, numpy/pandas/polars/pyarrow out, no library required

Documentation

Full documentation at walnutgeek.github.io/lythonic.

About

Lightweight pythonic toolkit for building complex workflows.

Resources

Stars

9 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages