agi-app-pandas-execution

agi-app-pandas-execution publishes the execution_pandas_project AGILAB app as a self-contained PyPI payload. It is a small execution benchmark for the Pandas worker path with an explicit Cython/C-speedup kernel.

Purpose

Use this package to validate AGILAB manager/worker distribution with a deterministic tabular workload. It generates CSV partitions, processes them with Pandas, and writes reducer evidence that can be compared across run modes. The default typed_numeric kernel exposes a contiguous float64 hot loop so Cython mode has real scalar work to compile instead of wrapping Pandas calls.

Installed Project

The distribution name is agi-app-pandas-execution; the AGILAB project name is execution_pandas_project. The package exposes both execution_pandas and execution_pandas_project through the agilab.apps entry point group, so AgiEnv(app="execution_pandas_project") works without a monorepo checkout.

Install

pip install agi-app-pandas-execution

Most users get this package through agi-apps, agilab[ui], or agilab[examples]; direct installation is useful when validating one app package in isolation.

Run In AGILAB

Select execution_pandas_project, open ORCHESTRATE, then run Deploy scheduler & workers and RUN. Keep the default local settings for a first proof, or disable Cython when you specifically want to compare the pure Python worker path. Increase partitions or worker count when you want a stronger distribution check.

Expected Inputs

The default run creates its own deterministic CSV workload under shared storage. No external dataset, cloud account, notebook, or API key is required.

Expected Outputs

Workers write processed tabular outputs and the reducer writes a summary with row counts, source files, engine labels, score metrics, and dtype/kernel metadata.

Change One Thing

Switch the workload kernel or partition count, then compare the reducer summary from the previous run. The app should keep the same input contract while making execution-mode differences visible.

Scope

This is a synthetic execution-path example. It is useful for validating Pandas worker behavior, not for demonstrating a domain analytics product.