--- title: RAG vs. Vectorless RAG, clearly explained! | daily.dev description: Vectorless RAG replaces embedding-based similarity search with LLM reasoning over a document's table of contents. Instead of chunking and embedding text, a... canonical: https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: RAG vs. Vectorless RAG, clearly explained! | daily.dev og:description: Vectorless RAG replaces embedding-based similarity search with LLM reasoning over a document's table of contents. Instead of chunking and embedding text, a... og:url: https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb og:image: https://api.daily.dev/og/posts/fKqJhnSfb.png og:image:alt: RAG vs. Vectorless RAG, clearly explained! og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # RAG vs. Vectorless RAG, clearly explained! **[Daily Dose of Data Science \| Avi Chawla \| Substack](https://daily.dev/sources/dailydoseofds)** · 7 min read · 10 upvotes · 0 comments ## Summary Vectorless RAG replaces embedding-based similarity search with LLM reasoning over a document's table of contents. Instead of chunking and embedding text, a preprocessing step builds a tree index where each node holds a section title, page range, and summary. At query time the LLM reads the tree, picks likely sections, pulls their full text, and answers with page citations, reopening more nodes if evidence is thin. On 63 questions over NVIDIA's FY2022 to FY2026 annual reports, tree retrieval answered all 63 correctly at about $110 per 1,000 queries, while the best vector setup hit 92.1% at $134, and top-10 vector search cost $15 but topped out at 76.2% accuracy. Trees suit long structured documents like filings and contracts; vector DBs still fit lots of short, unrelated text better. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is vectorless RAG and how does it differ from standard RAG? Vectorless RAG drops the embedding model and vector database, replacing similarity search with LLM reasoning over a document's structure. A preprocessing step builds a table-of-contents tree where each node has a section title, page range, and summary. At query time, the LLM reads the tree, selects likely sections, pulls their full text, and answers with page citations, reopening more nodes if the evidence is insufficient. _Weighing tree-based retrieval against vector search? daily.dev surfaces RAG architecture comparisons like this as they appear._ ### Why does standard RAG with vector search sometimes retrieve the wrong year's data from annual reports? Five annual reports from the same company usually follow an identical template with only the numbers changing, so a question about one fiscal year's chunk can embed just as close to a different year's chunk, since both chunks discuss the same topic. The LLM then answers from the wrong year because the retrieved chunk is genuinely on-topic, just from the wrong document. _Debugging retrieval accuracy on structured documents is easier with daily.dev tracking RAG failure patterns like this._ ### How much does tree-based retrieval cost compared to vector search for RAG over financial documents? On 63 questions over NVIDIA's FY2022 to FY2026 annual reports, tree retrieval answered all 63 correctly at about $110 per 1,000 queries. The best vector setup reached only 92.1% accuracy by sending the top 100 chunks at about $134 per 1,000 queries, while top-10 vector search cost about $15 but no embedding model exceeded 76.2% accuracy. _Choosing between retrieval architectures on cost and accuracy is a recurring decision daily.dev helps developers stay informed on._ ## Similar posts on daily.dev - [Proxy-Pointer RAG: Achieving Vectorless Accuracy at Vector RAG Scale and Cost](https://daily.dev/posts/proxy-pointer-rag-achieving-vectorless-accuracy-at-vector-rag-scale-and-cost-robiobrdq) · Towards Data Science · 1 upvotes · 0 comments - [I Stopped Using Vector Databases. My RAG Got Better.](https://daily.dev/posts/i-stopped-using-vector-databases-my-rag-got-better--msjc5zpqk) · Medium · 0 upvotes · 0 comments - [What RAG Actually Costs to Run in Production: A Full Cost Breakdown](https://daily.dev/posts/what-rag-actually-costs-to-run-in-production-a-full-cost-breakdown-26v3uoonx) · DigitalOcean Community · 0 upvotes · 0 comments - [RAG indexing: Structure and evaluate for grounded LLM answers](https://daily.dev/posts/rag-indexing-structure-and-evaluate-for-grounded-llm-answers-xf52bq2jy) · Meilisearch · 5 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#rag](https://daily.dev/tags/rag), [#vector-search](https://daily.dev/tags/vector-search), [#ai-inference](https://daily.dev/tags/ai-inference) [View this post on daily.dev](https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"RAG vs. Vectorless RAG, clearly explained!","url":"https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb"},"datePublished":"2026-10-30T20:27:54.731Z","dateModified":"2026-10-30T20:28:23.893Z","description":"Vectorless RAG replaces embedding-based similarity search with LLM reasoning over a document's table of contents. Instead of chunking and embedding text, a...","image":"https://media.daily.dev/image/fetch/s---6hkUNx4--/f_auto,pg_1,q_auto/https://substackcdn.com/image/fetch/%24s_%21I9LO%21%2Cw_1200%2Ch_675%2Cc_fill%2Cf_jpg%2Cq_auto:good%2Cfl_progressive:steep%2Cg_auto/https%253A%252F%252Fsubstack-post-media.s3.amazonaws.com%252Fpublic%252Fimages%252Fa86e8c9a-f522-4db7-8663-e5a94aec7689_1080x1080.gif?_a=BAMAMiiw0","thumbnailUrl":"https://media.daily.dev/image/fetch/s---6hkUNx4--/f_auto,pg_1,q_auto/https://substackcdn.com/image/fetch/%24s_%21I9LO%21%2Cw_1200%2Ch_675%2Cc_fill%2Cf_jpg%2Cq_auto:good%2Cfl_progressive:steep%2Cg_auto/https%253A%252F%252Fsubstack-post-media.s3.amazonaws.com%252Fpublic%252Fimages%252Fa86e8c9a-f522-4db7-8663-e5a94aec7689_1080x1080.gif?_a=BAMAMiiw0","isAccessibleForFree":true,"articleSection":"Daily Dose of Data Science | Avi Chawla | Substack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Daily Dose of Data Science | Avi Chawla | Substack","logo":"https://media.daily.dev/image/upload/s--4IHQgTOw--/f_auto/v1710503712/logos/dailydoseofds","url":"https://daily.dev/sources/dailydoseofds"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":9},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,rag,vector-search,ai-inference","timeRequired":"PT7M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Daily Dose of Data Science | Avi Chawla | Substack","item":"https://daily.dev/sources/dailydoseofds"},{"@type":"ListItem","position":3,"name":"RAG vs. Vectorless RAG, clearly explained!"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/rag-vs-vectorless-rag-clearly-explained--fkqjhnsfb#faq","mainEntity":[{"@type":"Question","name":"What is vectorless RAG and how does it differ from standard RAG?","acceptedAnswer":{"@type":"Answer","text":"Vectorless RAG drops the embedding model and vector database, replacing similarity search with LLM reasoning over a document's structure. A preprocessing step builds a table-of-contents tree where each node has a section title, page range, and summary. At query time, the LLM reads the tree, selects likely sections, pulls their full text, and answers with page citations, reopening more nodes if the evidence is insufficient. Weighing tree-based retrieval against vector search? daily.dev surfaces RAG architecture comparisons like this as they appear."}},{"@type":"Question","name":"Why does standard RAG with vector search sometimes retrieve the wrong year's data from annual reports?","acceptedAnswer":{"@type":"Answer","text":"Five annual reports from the same company usually follow an identical template with only the numbers changing, so a question about one fiscal year's chunk can embed just as close to a different year's chunk, since both chunks discuss the same topic. The LLM then answers from the wrong year because the retrieved chunk is genuinely on-topic, just from the wrong document. Debugging retrieval accuracy on structured documents is easier with daily.dev tracking RAG failure patterns like this."}},{"@type":"Question","name":"How much does tree-based retrieval cost compared to vector search for RAG over financial documents?","acceptedAnswer":{"@type":"Answer","text":"On 63 questions over NVIDIA's FY2022 to FY2026 annual reports, tree retrieval answered all 63 correctly at about $110 per 1,000 queries. The best vector setup reached only 92.1% accuracy by sending the top 100 chunks at about $134 per 1,000 queries, while top-10 vector search cost about $15 but no embedding model exceeded 76.2% accuracy. Choosing between retrieval architectures on cost and accuracy is a recurring decision daily.dev helps developers stay informed on."}}]} ```