What problem does it solve? Aggregating results from dozens of search engines, deduplicating them, and ranking them consistently is tedious to build from scratch, and individual scraper-based engines silently break or get skipped without visibility. This Skill guides you through the search-web-api package so you can run meta-searches, diagnose failing engines, and wire search into agents, Workers, or an MCP server. ## Core Features & Use Cases - Meta-search across 71 engines: Query engines across 13 categories (general, academic, it, news, images, videos, and more) in parallel, with automatic deduplication and weighted scoring via the ResultContainer. - Engine and category selection: Restrict searches by exact engine names or categories, inspect the engine catalogue, and check per-engine health status when results go missing. - Autocomplete and fallbacks: Use autocomplete backends (Google, DuckDuckGo, Brave, and others), a local transformers.js next-word model, or hosted fallbacks via Tavily and public SearXNG instances. - Use Case: You are building a research agent that needs academic sources. Call search.search(query, 1, undefined, ["academic"]) to fan out to arXiv, PubMed, Semantic Scholar, and Google Scholar, then receive one merged, score-ranked result list. ## Quick Start Ask the assistant to search the academic category for a topic like "attention is all you need" using the search-web-api package and return the top merged results.
