Concepts
Nyckel is built around a small set of ideas. Understanding them makes every other part of the product easier to reason about.
Input→Prediction + confidence→Review & feedback→Improved endpoint
Prediction endpoints
What they are, how they work, and why they differ from static models.
Confidence scores
What the score means, how to use it, and when to act on it.
Labels
How to define the decision space — the set of categories your function can predict.
Samples
Labeled examples that teach the model what each label means.
The feedback loop
How predictions improve over time as your application provides feedback.