Solid roundup! Having Tavily, Firecrawl, and Exa alongside inference providers like Groq and OpenRouter really highlights how much the modern AI stack relies on rich context retrieval (RAG) just as much as pure LLM capability.

Cohere Rerank and Jina Reader (r.jina.ai) in particular are absolute lifesavers for keeping prompt noise low and context relevance high. Thanks for putting together the clear code snippets for each.