wfc-client 0.1.0

Pure-stdlib Tier-1 sugar for writing Workflow Canvas (wfc) methods

wfc-client

Pure-stdlib Tier-1 sugar for writing Workflow Canvas (wfc) methods.

wfc-client is the opt-in ergonomic way to write a wfc method. The canonical interface is the Tier-2 env-var + file contract — wfc-client is a thin, zero-dependency wrapper over it. It never installs wfc, pandas, or any third-party package into your environment, and it never copies, reads, or serializes your data bytes: it is a metadata recorder.

import wfc_client as wfc

@wfc.method
def qc(ctx):
    clean_path = ctx.workdir / "clean.csv"
    # ... write your file with whatever library you like ...
    ctx.save_artifact("clean", clean_path)
    ctx.log_metric("kept_rows", 100)

if __name__ == "__main__":
    wfc.run()

The ctx surface

Member	Purpose
ctx.input(slot)	Resolved input paths (list[Path]) for an input slot.
ctx.params	Parsed params dict.
ctx.workdir	Scratch dir at WFC_RUN_DIR/_workdir/ (auto-created).
ctx.run_dir	WFC_RUN_DIR (advanced).
ctx.save_artifact(name, path)	Record that path is the declared output name. Path only.
ctx.log_metric(name, value)	Record a scalar metric.

At wfc.run() exit, wfc-client writes one _wfc_results.json manifest of {outputs, metrics} with run-dir-relative paths. The host reads it.

Exactly one @wfc.method per method module is required.