Project

routerlab

Route LLM steps by cost and quality, then keep multi-step agent chains inside a running task budget.

GitHub Repository All Projects

Focus

  • Model routing
  • Evaluation
  • AI infrastructure

GitHub Description

Cost-quality routing and budget-aware control for LLM workflows.

  • Focuses on provider-aware model selection, routing behavior, and budget-aware loop control.
  • Built around evaluation harnesses and runtime budget primitives rather than one-off manual checks.

Package Links

@routerlab/core Cost-quality router with budget-aware multi-step agent control. @routerlab/cli CLI for cost-quality routing experiments.

Preview / Demo

Routing preview

Use RouterLab to compare candidate models by cost, quality, task fit, and remaining chain budget before hard-coding provider choices into an application.

View RouterLab repository

Working Preview

Compare cost and quality before choosing a model route.

This static preview mirrors the kind of tradeoff table RouterLab is designed to produce for task-specific model routing and budget-aware agent-loop control.

claude-haiku Quality 78 $0.18 Fast draft
gpt-4o-mini Quality 82 $0.24 Balanced
claude-sonnet Quality 92 $1.20 High accuracy