Nyckel Documentation

Nyckel turns machine learning into a managed REST API. You define a function — classification, Box Detect, or Search — and Nyckel handles model selection, training, hosting, monitoring, and continuous improvement from real-world feedback.

Choose your path

Browse all documentation

A complete index of every page in the Nyckel documentation, grouped by section. This list is the authoritative sitemap for the /docs/ corpus.

Get Started

Welcome to Nyckel

• What is a Function?
• How Functions Work
• Function Types
• How Functions Improve from Feedback
• Build and Improve a Spam Classifier

Start with a Prebuilt Function

• Prebuilt Classification Functions

Create Your Own Classification Function

• Create a Function
• Define Labels
• Invoke Predictions
• Review and Improve Predictions
• Next Steps

Create Your Own Box Detect Function

• Create a Function
• Add Training Images and Boxes
• Invoke Detection
• Next Steps

Create Your Own Search Function

• Create a Function
• Add Reference Data
• Query Similar Content
• Next Steps

Where to Go Next

• Learn Core Concepts
• Improve Accuracy
• Build Production Workflows
• Explore the API Reference

Learn

• Overview

Concepts

• Overview
• Functions
• Models
• Labels
• Samples
• Confidence Scores
• The Feedback Loop
• Prediction Endpoints

Best Practices

• Overview
• Review and improve predictions
• Get reliable predictions
• Improve accuracy over time
• Confidence scores and thresholds

Developer Platform

• Overview

Architecture

• Principles
• Prediction endpoints
• Build a feedback loop
• Production integration patterns

API Reference

Also useful