KMeans

Overview

kmeans is a selection algorithm for routes that rely on a cluster-based selector.

It aligns to config/algorithm/selection/kmeans.yaml.

Key Advantages

• Keeps cluster-based selection separate from decision rules.
• Minimal route config when clustering is handled by the selector itself.
• Works well when workloads naturally group into prompt clusters.

What Problem Does It Solve?

When prompt traffic falls into a few recurring clusters, a cluster-based selector can choose the best model more cleanly than a hand-written priority rule. kmeans exposes that selector at the route level.

When to Use

• you have a cluster-based selector for candidate models
• prompt traffic naturally groups into repeatable classes
• the route should use learned clusters instead of static ranking

Configuration

Use this fragment inside routing.decisions[].algorithm:

algorithm:
  type: kmeans