exa-performance-tuning

# Exa Performance Tuning ## Overview Optimize Exa search API response times for production workloads. Key levers: search type selection (instant < fast < auto < neural < deep), result count reduction, content scope control, result caching, and parallel query execution. ## Latency by Search Type | Type | Typical Latency | Use Case | |------|----------------|----------| | `instant` | < 150ms | Real-time autocomplete, typeahead | | `fast` | p50 < 425ms | Speed-critical user-facing search | | `auto` | 300-1500ms | General purpose (default) | | `neural` | 500-2000ms | Best semantic quality | | `deep` | 2-5s | Maximum coverage, light deep search | | `deep-reasoning` | 5-15s | Complex research questions | ## Instructions ### Step 1: Match Search Type to Latency Budget ```typescript import Exa from "exa-js"; const exa = new Exa(process.env.EXA_API_KEY); function selectSearchType(latencyBudgetMs: number) { if (latencyBudgetMs < 200) return "instant"; if (latencyBudgetMs < 500) return "fast"; if (latencyBudgetMs < 1500) return "auto"; if (latencyBudgetMs < 3000) return "neural"; return "deep"; } async function optimizedSearch(query: string, latencyBudgetMs: number) { const type = selectSearchType(latencyBudgetMs); const numResults = latencyBudgetMs < 500 ? 3 : latencyBudgetMs < 2000 ? 5 : 10; return exa.search(query, { type, numResults }); } ``` ### Step 2: Minimize Content Retrieval ```typescript // Each content option adds latency. Only request what you need...