Skip to main content
Loading navigation
Back to challenges

Multi-Agent Hardware Cost and Latency Optimizer with CrewAI and OpenRouter

Orchestrate multi-model workflows to dynamic accelerator hardware tiers for cost optimization. Build a collaborative multi-agent system in CrewAI using OpenRouter API routing to evaluate query complexity and redirect traffic between high-tier and low-tier hardware targets.

EngineeringHosted by Vera
Sign in free to start

What you are building

Construct a CrewAI agent team using OpenRouter to evaluate prompt difficulty and route requests to optimal model hardware endpoints, minimizing token execution costs.

How work is evaluated

Evaluation

Evaluates task delegation decision quality and total cost efficiency of routed queries.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

Cost Model Selection Test

Verify cheap model selection for simple prompt.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

Routing Accuracy Rate

Percentage of requests routed to appropriate tier. • target: 0.95 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Orchestrate CrewAI role-playing agents for intent classification and hardware target selection.

  • Utilize OpenRouter API to route requests based on dynamic latency and cost limits.

  • Achieve over 30% cost reduction without sacrificing task accuracy scores.

Resources

Dataset notes

Dataset of 100 mixed complexity prompts ranging from simple math to complex code refactoring.

Challenge setup

Evaluates task delegation decision quality and total cost efficiency of routed queries.

Challenge input

JSON prompt text and complexity classification ground truth

CrewAI

Framework for role-playing multi-agent systems.

OpenRouter

Unified API gateway for dynamic model access.

Evaluated output

JSON selected model slug, estimated query cost

Evaluation checks
  • Verify cheap model selection for simple prompt.
  • Percentage of requests routed to appropriate tier. • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: LLM Routing Cost Efficiency Index
  • Routing Accuracy Rate target: 95%
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • CrewAI · Required
  • crewAI · Optional
Policy Serving
  • OpenRouter · Optional
Orchestration
  • OpenRouter · Optional

Evaluation contract

  • Cost Model Selection Test · Weight 1
  • Routing Accuracy Rate · Weight 1

Metrics

  • Routing Accuracy Rate · Target 0.95

Benchmarks

  • LLM Routing Cost Efficiency Index
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON prompt text and complexity classification ground truth

    Output: JSON selected model slug, estimated query cost

    Execution harness: Python

Evaluation tools

  • CrewAI · Agent Orchestration
  • OpenRouter · Model Gateway

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Multi-Agent Hardware Cost and Latency Optimizer with CrewAI and OpenRouter