Open sourcePython

Giskard

One command that hunts for the problems in your AI you did not think to test.

What it is

Giskard is a free Python library that scans an AI app for weaknesses automatically. You wrap your app, run one scan, and it probes for hallucination, bias, prompt injection and other issues, then hands you an HTML report.

Why a QA engineer cares

It is like an automated security and quality scan for AI. Instead of thinking up every bad input yourself, Giskard generates them and tells you where the app breaks, which is a fast way to find risks before launch.

Get started

Install it, then run the example below.

Install

pip install giskard

Quickstart

import giskard

# wrap your AI app as a Giskard model, then scan it
model = giskard.Model(model=my_app, model_type="text_generation",
                      name="Support bot", description="Answers refund questions")

scan_results = giskard.scan(model)
scan_results.to_html("giskard_report.html")   # open the report

What you get

✓Automatic scan for hallucination, bias and injection
✓Clear HTML report of what broke and why
✓Turn found issues into a reusable test suite
✓Works for LLM apps and classic ML models

Best for

Pre-launch risk scanSafety and biasVulnerability hunting

Where it fits

Run it before release; keep DeepEval or promptfoo for your everyday regression evals.

Official docs for Giskard →

Other AI-testing tools

promptfoo

Run evals and safety tests on prompts and models, works in CI

DeepEval

Pytest-style AI evals (correct, relevant, bias)

Ragas

Check RAG bots: did they find and use the right source

LangSmith

Trace, save datasets and run evals for AI apps

Confident AI

Dashboards to track eval scores over time

Put it to work

See where Giskard fits in the full picture, or follow the step-by-step AI for QA roadmap.

The AI testing guide The AI for QA roadmap