Community guidelines
Be specific and constructive. No vendor spam — promoting your own product belongs in a listing. Anyone can read; posting needs a free account.
our vendor says we can skip vector search and just put the relevant project documents into a large context window. That sounds much simpler than chunking, embeddings and retrieval. Has anyone replaced RAG this way for a real knowledge base?
Large context is great when the document set is small and you already know which files matter. It doesnt solve finding those files, access control, freshness or cost. We use long context after retrieval, not instead of it
test it with questions that need one sentence buried in twenty documents. Models can accept a huge prompt without using every part equally well. Also calculate the price and latency at your normal traffic, not for one demo
our set is around 60,000 documents with permissions by department, so i guess the word relevant is doing a lot of work in the vendor pitch
Exactly. You still need a retrieval and permission layer. Bigger context lets you send richer evidence once you find it. It can reduce aggressive chunking and multi step calls, but it doesnt turn a document warehouse into an answer system by itself