This repository contains eqty.sdk.r, an R package that makes the
eqty-sdk Python SDK available to R
users. It also contains examples and sample Integrity Manifests to help users
get started quickly.
EQTY SDK for R is intended for data scientists, statisticians, ML engineers, and teams that need a trustworthy record of how an output was produced. It adds provenance to an R workflow without requiring its analytical logic to be rewritten.
It creates a clear record of the data, code, models, settings, and processing steps behind each result. Teams can use that record to understand what produced an output, reproduce a previous run, compare results, and investigate unexpected changes. It also provides evidence that can be shared with reviewers, collaborators, and auditors without relying on incomplete notes or manual reconstruction of the workflow.
The provenance engine, asset models, content identifiers, signing, and manifest
operations are implemented by the Python SDK. The R package loads that
implementation through reticulate and exposes its classes and functions with
R-accessible names and helpers. R users can therefore create assets, record
computations, sign lineage statements, and export Integrity Manifests without
writing Python code.
| Path | Contents |
|---|---|
package/ |
R package source, package documentation, metadata, and release notes |
samples/ |
Examples from basic lineage to multi-stage agent workflows |
samples/manifests/ |
Integrity Manifests that can be opened in EQTY Explorer |
skills/ |
Installable agent skills for using and instrumenting R workflows with eqty.sdk.r |
- Read the R package guide for requirements, installation, Python dependency management, API usage, and package development.
- Follow the samples guide for runnable workflows and pre-generated Integrity Manifests. Open the manifests in EQTY Explorer to inspect assets, metadata, computations, and their connections, then compare each graph with the R code that created it.
- Use the CRAN package page to install
the released
eqty.sdk.rpackage. - For R code, use the R package guide, generated R help, and R samples. The Python SDK is an implementation detail of the R package.
Install the eqty-sdk-r skill for Codex, Claude Code, OpenCode, or another
supported coding agent with the skills CLI:
npx skills add https://github.com/eqtylab/integrity-r --skill eqty-sdk-rThe skill teaches an agent how to register R assets and computations, manage
contexts and signers, handle _store safely, and export local Integrity
Manifests. Its source is
skills/eqty-sdk-r/SKILL.md.
By default, instrumentation is written to <source-stem>-instrumented.R with a
companion <source-stem>-instrumented.md lineage mapping.
The command installs only the eqty-sdk-r skill directory, including its
bundled references. It does not copy this repository's root AGENTS.md, package
guide, or sample folders into the target project. The installed skill is
self-contained and directs agents to use the R API without Python imports or
Python reflection.
Installing the skill does not install eqty.sdk.r or its Python backend. The R
package declares eqty-sdk through reticulate::py_require(), so reticulate may
download the managed Python dependency when an EQTY workflow first uses the
SDK. Planning and syntax-only validation should not trigger that provisioning.
The R package source lives under package/, and its reference documentation is
generated from roxygen comments during the release build.
Run./build-package.sh to regenerate the documentation, build the source tarball,
and perform the CRAN checks. The script stops if documentation generation or
packaging fails, or if R CMD check reports an error or warning. Successful
build and check results are written under dist/.
Pushing a version tag that matches the version in package/DESCRIPTION runs
the release workflow and attaches the checked source tarball to a GitHub Release.
See the package guide for details.
See CONTRIBUTING.md for the current contribution policy. Use the GitHub issue forms for bugs, questions, and feature requests.
Report suspected vulnerabilities privately according to SECURITY.md.
This project is licensed under the Apache License 2.0.