Developer Shares Practice of Building RAG Search for Gaming Site Using Only Cloudflare

On November 11, 2026, a developer writing on DEV Community shared how they built a RAG-powered search and Q&A experience for a retro game site using only Cloudflare. The platform, called NESBox, is an online multiplayer site for retro games. Its library holds over a hundred titles, with descriptions written in a mix of Chinese, English and Japanese. Because players frequently cannot recall exact titles and search with vague phrases instead, such as "that off-road motorbike one" or "a fighting game two people can play together", the developer added natural-language retrieval to the site.

The architecture is deliberately serverless. Rather than running a standalone backend or an external vector database, the developer combined Cloudflare Workers, Workers AI, Vectorize and KV so that a single Worker handles content ingestion, hybrid retrieval and streaming Q&A, all within one Worker file. Workers AI performs the embedding and generation work, Vectorize provides the vector index, and KV stores the underlying data.

Later that same morning, at 09:49, the Chinese developer community Juejin republished an introduction to the same project, titled "Building RAG Search and Q&A for a Game Site Using Only the Cloudflare Suite". The post describes the identical stack, noting that no server and no separate vector database are needed: Workers, Workers AI, Vectorize and KV together implement ingestion, hybrid retrieval and streaming Q&A inside a single Worker file. Neither piece goes beyond the technical description, and neither reports launch dates, traffic figures or future plans, so as of November 11, 2026 the project is publicly known only through these write-ups.