robusto-cloud 0.1.1

Python client for Robusto Cloud — adversarial robustness testing and hardening for ML models

Robusto

Adversarial robustness hardening for ML models.

Upload your pre-trained PyTorch model, Robusto stress-tests it against adversarial attacks and returns a hardened version with a compliance report.

pip install robusto-cloud → robusto configure → robusto init → hardened model + report

Quick Start

pip install robusto-cloud
robusto configure          # save your API key
robusto init               # interactive: pick model, dataset, domain, attacks → submit

Or scripted:

robusto harden -m model.pt -d data.csv --domain tabular --mode standard
robusto watch <job_id>
robusto download <job_id>

Or in Python:

from robusto import Robusto

client = Robusto()
job = client.harden(model_path="model.pt", dataset_path="data.csv", domain="tabular")
result = client.wait(job["job_id"])
client.download_model(job["job_id"], "hardened_model.pt")
client.download_report(job["job_id"], "report.pdf")

Domains

Image, Tabular, Graph, Point Cloud, Sensor

Modes

Mode	Use case
low	Quick audit
standard	Balanced (recommended)
rigorous	Full compliance

Documentation

• SDK Guide — install, configure, CLI + Python usage, full API reference
• API Reference — REST endpoints, auth, error codes
• Compliance Reports — NIST AI RMF + EU AI Act mapping

Running Locally (contributors)

git clone https://github.com/Cortexa-Labs-Inc/Cortexa-Labs-Robusto.git
cd Cortexa-Labs-Robusto
pip install -r requirements.txt && pip install -e .
export ROBUSTO_API_KEY=my-local-key
python -m robusto.server

Then point the SDK at localhost:

client = Robusto(api_key="my-local-key", base_url="http://localhost:8000/api/v1")

License

Proprietary. Copyright 2026 Cortexa Labs.