SkyAIAppRouting Strategies Guide
Master routing strategies to achieve optimal balance of cost, quality, and performance.
Strategy Overview
balancedBalancedBalance cost, quality, and latency
Best for
General useProductionMost applications
Metrics
Cost
Quality
Speed
cost-optimizedCost-OptimizedMinimize API call costs
Best for
Batch processingData labelingSimple tasks
Metrics
Cost
Quality
Speed
quality-firstQuality-FirstPrioritize most powerful models
Best for
Complex reasoningCreative generationCritical decisions
Metrics
Cost
Quality
Speed
latency-optimizedLatency-OptimizedChoose fastest responding models
Best for
Real-time chatUser interfacesInstant feedback
Metrics
Cost
Quality
Speed
Custom Strategies
Create custom strategies with precise control over model selection logic:
const response = await sky.route({
goal: "cost",
strategy: {
name: "custom-strategy",
rules: [
{
condition: { complexity: "low" },
models: ["gpt-5-nano", "claude-haiku-4.5"],
weight: { cost: 0.8, quality: 0.2 }
},
{
condition: { complexity: "high" },
models: ["gpt-5.2-thinking", "claude-opus-4.5"],
weight: { cost: 0.3, quality: 0.7 }
}
],
fallback: ["gpt-5-mini"],
cache: true
},
messages: [...]
});Best Practices
Choose strategy by scenario
Use cost-optimized for batch, latency-optimized for real-time
Enable semantic caching
Save 30-40% costs, especially for repetitive queries
Configure fallback chains
Ensure high availability with automatic model switching
Monitor and optimize
Regularly review analytics and tune strategy parameters
Was this page helpful?
Let us know how we can improve