file-manager is a Codex plugin that normalizes dataset files under a user-specified data directory. When the user asks to download, process, or analyze a dataset, the plugin keeps raw files, processed files, scripts, logs, and analysis results in a predictable single- or multi-dataset layout instead of scattering them across the working directory.
Requested analyses live beside raw/ and processed/ under data_analysis/<user-chosen-name>/. Each analysis directory contains only analysis.py and results.txt unless the user explicitly asks for additional outputs.
When an existing processed dataset needs filtering or refinement, update the existing processed/process.py and safely replace the canonical processed output. Do not create parallel filter scripts or duplicate output variants. Record the exact rules, parameters, before-and-after counts, removals by reason, validation results, output replacement, and any errors in the existing processed/process.log.
This repository is public and can be installed directly from the GitHub marketplace source.
Add the repository as a Codex plugin marketplace:
codex plugin marketplace add promelin/file_manager
codex plugin add file-manager@file-managerAlternatively, after adding the marketplace, restart the ChatGPT desktop app, open the Plugins Directory, select File Manager, and install the plugin there.
Describe a model-development task naturally or invoke the bundled skill explicitly as $file-manager.
Examples:
Use $file-manager to download and preprocess CIFAR-10 for this training project.Organize MNIST and Fashion-MNIST as separate datasets for model evaluation.Analyze solvent kinds in the managed dataset and use solvent_kinds as the analysis directory.Create the data workflow, but do not download or process anything yet.