Quickstart
Take one question through a complete Knotest workflow.
Begin with a question you can test: for example, whether an image model remains useful when image quality changes. This is an illustrative exercise; choose inputs and criteria appropriate to your system.
Create a Project
Open Knotest and sign in, then use the top-bar Project menu or ⌘/Ctrl+O to create a Project. Give it a name that describes the system or question.
Prepare the inputs
In Vault, import a Dataset and inspect its task, samples, and reference labels. Add a compatible Model or verified API profile. Keep their versions identifiable.
Save a Test plan
Open Testing and create a Test plan. State the question, recipient, requirements, and reason for each test. A baseline gathers information; an acceptance decision needs confirmed criteria.
Review and run
Prepare a Run. Review its inputs, tests, criteria, and assumptions. Complete any inference setup and license consent, then choose Run tests when ready.
Review the result
Inspect Scoring and the supporting examples. Open Report to read the question, measurements, limits, and source basis, then export a PDF if needed.
Put the result in context
If the work informs a decision about an AI system, create an Assessment, link the result to its requirement, and record a human conclusion. The decision retains the reviewed evidence and remaining gaps.