--- title: A navigable context layer for coding agents | daily.dev description: Splitting Redis documentation into three linked entities, pages, chunks, and commands, raised agent retrieval nDCG@5 from 0.449 to 0.619 on a 62-question test... canonical: https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: A navigable context layer for coding agents | daily.dev og:description: Splitting Redis documentation into three linked entities, pages, chunks, and commands, raised agent retrieval nDCG@5 from 0.449 to 0.619 on a 62-question test... og:url: https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil og:image: https://api.daily.dev/og/posts/t8PQdBLIl.png og:image:alt: A navigable context layer for coding agents 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. # A navigable context layer for coding agents **[Redis](https://daily.dev/sources/redislabs)** · 15 min read · 0 upvotes · 0 comments ## Summary Splitting Redis documentation into three linked entities, pages, chunks, and commands, raised agent retrieval nDCG@5 from 0.449 to 0.619 on a 62-question test set. The original design used one broad entity, RedisioDoc, which mixed page summaries, passages, and commands into the same search, causing failures like BF.RESERVE never appearing in results. The new version connects entities through a shared page_id so the agent can move from discovery to evidence. The gains came with a cost: median tool calls rose from 2 to 4, median latency rose from 6.8s to 10.5s, and p95 latency hit 17.9s, consuming 72% of the 25-second request budget. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I improve RAG retrieval accuracy for an AI coding agent answering documentation questions? Splitting documentation into separate linked entities for pages, chunks, and commands, rather than one broad document type, improved retrieval nDCG@5 from 0.449 to 0.619 in one evaluation. The entities were connected via a shared page_id so an agent could discover a page, then fetch chunks or structured command references from it, rather than searching one undifferentiated index. _Anyone tuning RAG pipelines for coding agents can track architecture lessons like this on daily.dev._ ### What metric should I use to evaluate retrieval quality for a RAG system besides recall? nDCG@5 (normalized discounted cumulative gain) rewards placing the correct result near the top, not just finding it somewhere in the results, and weights primary sources more than related ones. Recall only checks whether an expected page appears anywhere in the result set, which misses the fact that an agent reading results in order is unlikely to use a correct page buried at position nine. _Developers picking retrieval metrics for their own RAG evals can keep up with approaches like this on daily.dev._ ### What are the tradeoffs of adding more structured entities to a retrieval-augmented agent system? Adding more specific entities (pages, chunks, commands) with explicit relationships improved retrieval ranking across every question group, but increased agent tool calls and tail latency. Median tool calls rose from 2 to 4 and p95 latency rose from about 10 seconds to nearly 18 seconds, consuming 72% of a 25-second request limit versus 40% before. _Teams weighing richer retrieval schemas against latency budgets can follow this tradeoff via daily.dev._ ## Similar posts on daily.dev - [How we built the Redis Docs MCP for agents](https://daily.dev/posts/how-we-built-the-redis-docs-mcp-for-agents-rblycygla) · Redis · 6 upvotes · 1 comments - [Context Engineering for AI Agents Starts with Retrieval](https://daily.dev/posts/context-engineering-for-ai-agents-starts-with-retrieval-or6sfdhgx) · LangChain · 1 upvotes · 0 comments - [3x Faster Search: Parallel Test-Time Scaling with Instructed-Retriever-1](https://daily.dev/posts/3x-faster-search-parallel-test-time-scaling-with-instructed-retriever-1-mzfvguqym) · databricks · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#mcp](https://daily.dev/tags/mcp), [#rag](https://daily.dev/tags/rag), [#redis](https://daily.dev/tags/redis), [#vector-search](https://daily.dev/tags/vector-search) [View this post on daily.dev](https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil) ```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":"A navigable context layer for coding agents","url":"https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil"},"datePublished":"2026-10-31T00:26:28.087Z","dateModified":"2026-10-31T00:27:01.722Z","description":"Splitting Redis documentation into three linked entities, pages, chunks, and commands, raised agent retrieval nDCG@5 from 0.449 to 0.619 on a 62-question test...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f49038f52fd3bf108b764f76bfcd84fe?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f49038f52fd3bf108b764f76bfcd84fe?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Redis","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":"Redis","logo":"https://media.daily.dev/image/upload/s--hr1KV229--/f_auto/v1729434313/logos/redis.png","url":"https://daily.dev/sources/redislabs"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,mcp,rag,redis,vector-search","timeRequired":"PT15M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Redis","item":"https://daily.dev/sources/redislabs"},{"@type":"ListItem","position":3,"name":"A navigable context layer for coding agents"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/a-navigable-context-layer-for-coding-agents-t8pqdblil#faq","mainEntity":[{"@type":"Question","name":"How can I improve RAG retrieval accuracy for an AI coding agent answering documentation questions?","acceptedAnswer":{"@type":"Answer","text":"Splitting documentation into separate linked entities for pages, chunks, and commands, rather than one broad document type, improved retrieval nDCG@5 from 0.449 to 0.619 in one evaluation. The entities were connected via a shared page_id so an agent could discover a page, then fetch chunks or structured command references from it, rather than searching one undifferentiated index. Anyone tuning RAG pipelines for coding agents can track architecture lessons like this on daily.dev."}},{"@type":"Question","name":"What metric should I use to evaluate retrieval quality for a RAG system besides recall?","acceptedAnswer":{"@type":"Answer","text":"nDCG@5 (normalized discounted cumulative gain) rewards placing the correct result near the top, not just finding it somewhere in the results, and weights primary sources more than related ones. Recall only checks whether an expected page appears anywhere in the result set, which misses the fact that an agent reading results in order is unlikely to use a correct page buried at position nine. Developers picking retrieval metrics for their own RAG evals can keep up with approaches like this on daily.dev."}},{"@type":"Question","name":"What are the tradeoffs of adding more structured entities to a retrieval-augmented agent system?","acceptedAnswer":{"@type":"Answer","text":"Adding more specific entities (pages, chunks, commands) with explicit relationships improved retrieval ranking across every question group, but increased agent tool calls and tail latency. Median tool calls rose from 2 to 4 and p95 latency rose from about 10 seconds to nearly 18 seconds, consuming 72% of a 25-second request limit versus 40% before. Teams weighing richer retrieval schemas against latency budgets can follow this tradeoff via daily.dev."}}]} ```