exa-cost-tuning

# Exa Cost Tuning ## Overview Reduce Exa API costs through strategic search type selection, result caching, query deduplication, and usage monitoring. Exa charges per search request with costs varying by search type and content retrieval options. ## Cost Drivers | Factor | Higher Cost | Lower Cost | |--------|-------------|------------| | Search type | `deep-reasoning` > `deep` > `neural` | `keyword` < `fast` < `instant` | | numResults | 10-100 results | 3-5 results | | Content retrieval | Full text + highlights + summary | Metadata only (no content) | | Content length | `maxCharacters: 5000` | `maxCharacters: 500` | | Live crawling | `livecrawl: "always"` | Cached content (default) | ## Instructions ### Step 1: Match Search Config to Use Case ```typescript import Exa from "exa-js"; const exa = new Exa(process.env.EXA_API_KEY); // Define cost tiers per use case const SEARCH_PROFILES = { // Cheapest: metadata-only keyword search "autocomplete": { type: "instant" as const, numResults: 3 }, // Low cost: fast search with minimal content "quick-lookup": { type: "fast" as const, numResults: 3 }, // Medium: balanced search for RAG "rag-context": { type: "auto" as const, numResults: 5, text: { maxCharacters: 1000 }, }, // Higher cost: deep research "deep-research": { type: "neural" as const, numResults: 10, text: { maxCharacters: 3000 }, highlights: { maxCharacters: 500 }, }, }; async function costAwareSearch( query: string...