Fetch-once, serve-many scraping gateway for the AI-agent era: single-flight request coalescing + TTL cache with ETag/304 revalidation + HTML->markdown. Stdlib-only core; benchmarks show ~96% of redundant origin load removed and N-agent thundering herds collapsed to a single upstream fetch.
caching scalability gateway reverse-proxy web-scraping from-scratch ai-agents llm request-coalescing single-flight
-
Updated
May 30, 2026 - Python