Skip to main content

ADMET Workbench

Getting Started

pip install workbench
workbench
Workbench Version: 0.8.475
Bosco: Claude via Anthropic API key

Welcome to Workbench!

🐶  New to Workbench? Ask me to walk you through building your first model.

Workbench:Bosco>

No AWS account needed to start. Local mode gives you the whole chain — DataSource → FeatureSet → Model → Endpoint — on your filesystem, training with the same generated script SageMaker runs. pub_data reads public ADMET datasets straight from S3, so there's real data to model from the first prompt:

Workbench:Bosco> build me a solubility model from the aqsol public data

Ask Bosco in plain language, or type Python — both land in the same session, and the variables persist for both of you. See Local Mode.

Giving Bosco a model

Bosco is the resident ML agent. He needs a path to Claude, and status always names the one he took:

  • With an AWS account — he runs on Bedrock inside it, with Zero Data Retention. Nothing to install and nothing leaves your account.
  • Without one — set ANTHROPIC_API_KEY and restart. Bring your own key, or email workbench@supercowpowers.com and we'll issue you a trial key.

Connecting your AWS account

📝 Bosco: Workbench ML Agent — a real eleven-turn session, start to finish.

Chemprop Models

All the rage for the Open ADMET Challenge. ADMET Workbench supports:

  • Single Task Chemprop Models
  • Multi Task Chemprop Models
  • Chemprop + Descriptors Models (MPNN + Molecular Descriptors)
  • Foundation Chemprop Models (CheMeleon Pretrained)

Examples: Deploying Chemprop Models  ·  Deploying Foundation Chemprop Models

References

Chemprop Action Shots!

theme_dark
theme_quartz

The Dashboard

Health monitoring, dynamic updates, and a high-level summary across the whole pipeline, with drill-down views for Incoming Data, Glue Jobs, DataSources, FeatureSets, Models, and Endpoints. Try it: Workbench Dashboard Demo.

Private SaaS Architecture

Secure your Data, Empower your ML Pipelines

Workbench deploys as an AWS Stack inside your own cloud environment (BYOC), so your data never leaves it — and plugin support lets you tailor it to your own workflows. See the Private SaaS Architecture deck.

Documentation

The ADMET Workbench framework makes AWS® both easier to use and more powerful: a simple Python API over Glue, Athena, Feature Store, Models, and Endpoints, with web interfaces on top of it.

Workbench Docs covers the Python API in depth, fully searchable, with code examples throughout. Runnable versions of every example live in examples/.

Workbench takes something genuinely complex — the full set of AWS ML services — and makes it less complex, so there's conceptual documentation too: Workbench Presentations.

AWS Marketplace

Workbench is on the AWS Marketplace as a Dashboard for ML Pipelines and can be billed through AWS.

Installation extras

pip install workbench               # API + REPL + orchestration (the default)
pip install 'workbench[ui]'         # + plotly, dash — the Workbench Dashboard
pip install 'workbench[modeling]'   # + torch, chemprop, ray[tune], optuna
pip install 'workbench[misc]'       # + umap-learn
pip install 'workbench[dev]'        # + pytest, coverage, flake8, black
pip install 'workbench[all]'        # everything above

Quotes are needed — shells read square brackets as globs.

Cleanlab/Datalab workflows aren't bundled in any extra. Install cleanlab[datalab]>=2.8.0 before calling cleanlab_model().

Questions?

The SuperCowPowers team is happy to answer any questions about AWS and Workbench — workbench@supercowpowers.com or Discord.

Beta Program — if your company would like to be a Workbench Beta Tester, get in touch at the same address.

Contributions are welcome, and fall under the existing project license.

® Amazon Web Services, AWS, the Powered by AWS logo, are trademarks of Amazon.com, Inc. or its affiliates

Metadata

Release files for workbench 0.8.517

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for workbench 0.8.517
File Size Uploaded Provenance Actions
workbench-0.8.517.tar.gz 4.8 MB With provenance Details for workbench-0.8.517.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for workbench 0.8.517
File Interpreter ABI Platform Actions
workbench-0.8.517-py3-none-any.whl Python 3 none any Details for workbench-0.8.517-py3-none-any.whl

Total release size: 6.1 MB

Release files / workbench-0.8.517.tar.gz

Download URL workbench-0.8.517.tar.gz
Size 4.8 MB
Tags Source
SHA-256 checksum
How to use checksums
f72239a15626b94f85b174c215e6206112905545a192064075c09e5ff67d79f7
BLAKE2b-256 checksum
How to use checksums
988191b83655695cd983a7514756f52ab6ebe4fc3c0d05aefeec9face885d7a1
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.

Transparency log

Release files / workbench-0.8.517-py3-none-any.whl

Download URL workbench-0.8.517-py3-none-any.whl
Size 1.2 MB
Tags Python 3
SHA-256 checksum
How to use checksums
2b4fcc37cbe2f541c6c584f62dafe1cc4c10109e058517df9d15fc37ca6595d4
BLAKE2b-256 checksum
How to use checksums
dce5647a3ebc39febdba71cd1e9435f4af69a411e68096f10d1cd8b50976c11f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.16

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.

Transparency log

Security for workbench 0.8.517

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for workbench-0.8.517.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Nov 3, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.8.517 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page