IGC: Intelligence-Gated Crawling for Distributed Web Content Acquisition and RAG-Ready Vectorization

IGC: Intelligence-Gated Crawling for Distributed Web Content Acquisition and RAG-Ready Vectorization | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article IGC: Intelligence-Gated Crawling for Distributed Web Content Acquisition and RAG-Ready Vectorization Sharan Kumar Yenugula This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9039974/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Large-scale Retrieval-Augmented Generation (RAG) systems require high-quality web corpora, yet conventional crawlers optimise for coverage and throughput rather than semantic content quality, leaving the burden of corpus cleaning as an expensive post-processing step. We introduce IGC (Intelligence-Gated Crawler), a distributed web acquisition framework that integrates multidimensional content quality evaluation directly into the crawl pipeline, preventing low-value content from entering the embedding layer at source. IGC combines distributed crawling via BullMQ and Redis, structured HTML extraction through the Mozilla Readability algorithm, a six-dimensional content quality scoring model (length, density, readability, structure, uniqueness, freshness), semantic sentence-window chunking, and a pluggable embedding pipeline supporting OpenAI, Cohere, and locally-hosted models. Embeddings are persisted in PostgreSQL with the pgvector extension, enabling approximate nearest-neighbour cosine similarity retrieval. A pilot evaluation on commodity hardware (Intel Core i3, 8 GB RAM) demonstrates stable crawl throughput of approximately 5.4 pages per second with a mean latency of 1,241 ms, while the Intelligence-Gated Crawling model filters roughly 14% of crawled pages and reduces downstream embedding noise by approximately 25%. IGC provides a scalable data-ingestion layer for retrieval-augmented generation systems, semantic search engines, and large language model dataset construction pipelines. Information Retrieval and Management Artificial Intelligence and Machine Learning Web Crawling Distributed Systems Retrieval-Augmented Generation Semantic Search Vector Embeddings Content Extraction AI Data Pipelines Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9039974","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":601240169,"identity":"bb3d6dc2-caf1-40e5-87b7-f62897dbeb0d","order_by":0,"name":"Sharan Kumar Yenugula","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABNUlEQVRIie2RMUvDUBDHX3jw4vBa1wu1iR+h5UEoNOJXaRDS5WUSnEQKgXOqs+LgVwgIxbEQfF1C5wiCQ6GDWKgUpQ6KScClJOoomN/057gfd8cRUlHxd+kQAnlwONODcRpg5wcFvhTPrHPVyyr8t0okTJCtLJUqrclQGS8IlnUZzJdvN56LIFePyXGHEz26DYuUeOo1mgjt8F6Ji2HsuMgXo65U6WLc85IiJZF2w0DQQugJUsN0iu6PhGSpAtwuVB6ecmX/6ry/0t4xcpHIuZAf3ygJt41nBHeQSEFrGAm2JenMx3LFiKXokikchIk8pE30TMaVTf0z4Kzklvokbt+tj5y9dLFrbYEOt06D2Uq+npjbeqSKlN0xIXTzBSz/EStoz7AGhGjrjSJdlnRXVFRU/E8+Af+vYyoKJp1EAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0009-0008-5948-4608","institution":"Marri Laxman Reddy Institute of Technology and Management","correspondingAuthor":true,"prefix":"","firstName":"Sharan","middleName":"Kumar","lastName":"Yenugula","suffix":""}],"badges":[],"createdAt":"2026-03-05 12:26:48","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-9039974/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9039974/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104057916,"identity":"e8f86822-2782-45a5-a52c-1980816fb5df","added_by":"auto","created_at":"2026-03-06 09:06:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":69010,"visible":true,"origin":"","legend":"\u003cp\u003eHigh-level IGC architecture. (1) Scheduler injects seed URLs. (2) Redis/BullMQ queue distributes jobs. (3) Crawl Worker fetches and renders pages. (4) Content Extraction isolates article text. (5) Intelligence-Gated Crawling quality gate. (6) Semantic Chunking produces token-bounded\u003c/p\u003e","description":"","filename":"sysArch.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/93861ab8b9aee26197aecc81.png"},{"id":104057917,"identity":"941d5493-d943-402b-898e-98a79199fc67","added_by":"auto","created_at":"2026-03-06 09:06:58","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":42067,"visible":true,"origin":"","legend":"\u003cp\u003eSequence diagram of the IGC data pipeline: from job enqueue through distributed crawl, content extraction, Intelligence-Gated Crawling quality scoring, semantic chunking, embedding generation, and vector storage.\u003c/p\u003e","description":"","filename":"dataPipeline.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/b652844ebd7d9363eab14e08.png"},{"id":104057918,"identity":"ac6d3cb0-4e53-4cec-bee2-8accfb213e21","added_by":"auto","created_at":"2026-03-06 09:06:58","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":105781,"visible":true,"origin":"","legend":"\u003cp\u003ePer-page processing pipeline inside the IGC Crawler Worker: HTTP fetch, SPA detection and Playwright rendering fallback, Cheerio metadata extraction, JSDOM construction, Mozilla Readability article isolation, IntelligenceGated Crawling quality scoring, and conditional semantic chunk emission.\u003c/p\u003e","description":"","filename":"processPipeine.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/baa23f1baa828a08394a571d.png"},{"id":104402813,"identity":"715de1db-6fca-4504-8175-3fef004db0fe","added_by":"auto","created_at":"2026-03-11 12:16:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":18505,"visible":true,"origin":"","legend":"\u003cp\u003eExtractor Worker processing path: polling un-embedded chunks from PostgreSQL, dispatching batch embedding requests to the configured provider, detecting the pgvector column type, and persisting the resulting vectors.\u003c/p\u003e","description":"","filename":"extractorWorker.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/f4c97bbe3e29f9c4ba608eb6.png"},{"id":104057921,"identity":"3fe9931a-3335-492f-80f4-df76a0881dfc","added_by":"auto","created_at":"2026-03-06 09:06:58","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":25626,"visible":true,"origin":"","legend":"\u003cp\u003eIntelligence-Gated Crawling six-dimensional quality scoring model. Each sub-score is summed to a composite 0–100 score; the configurable quality gate blocks content below the threshold from proceeding to chunking and embedding.\u003c/p\u003e","description":"","filename":"qualityScoring.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/e8d9d72c7c6a009c7a047b7c.png"},{"id":104402418,"identity":"d10e4eb1-2836-4796-8640-783e38155b38","added_by":"auto","created_at":"2026-03-11 12:15:19","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":34674,"visible":true,"origin":"","legend":"\u003cp\u003eSemantic sentence-window chunking pipeline in IGC. Extracted text is segmented into paragraph blocks annotated with the nearest preceding section heading. Sentences are accumulated into token-bounded chunks (target: 512 tokens) with a two-sentence contextual overlap at each boundary\u003c/p\u003e","description":"","filename":"chunking.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/fdec7b246e80d3fbcf087bd0.png"},{"id":104057922,"identity":"069ebee3-f19a-47a3-9604-69168f3905ca","added_by":"auto","created_at":"2026-03-06 09:06:58","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":46775,"visible":true,"origin":"","legend":"\u003cp\u003eExtractor Worker embedding pipeline: DB polling loop, batch construction with section-heading context prefix, provider dispatch (OpenAI, Cohere, or local Ollama), pgvector column-type detection, and vector persistence.\u003c/p\u003e","description":"","filename":"embeddingPipeline.png","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1/08d447caf1c321f660e97a93.png"},{"id":104409032,"identity":"ae043bf8-8cb0-4257-b010-96fb91f2aaf7","added_by":"auto","created_at":"2026-03-11 12:44:00","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":654799,"visible":true,"origin":"","legend":"","description":"","filename":"IGCScript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9039974/v1_covered_34cb0165-7a77-4a1d-983f-64f66ca580ff.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eIGC: Intelligence-Gated Crawling for Distributed Web Content Acquisition and RAG-Ready Vectorization\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Marri Laman Reddy Institute of Technology and Management","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Web Crawling, Distributed Systems, Retrieval-Augmented Generation, Semantic Search, Vector Embeddings, Content Extraction, AI Data Pipelines","lastPublishedDoi":"10.21203/rs.3.rs-9039974/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9039974/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eLarge-scale Retrieval-Augmented Generation (RAG) systems require high-quality web corpora, yet conventional crawlers optimise for coverage and throughput rather than semantic content quality, leaving the burden of corpus cleaning as an expensive post-processing step. We introduce IGC (Intelligence-Gated Crawler), a distributed web acquisition framework that integrates multidimensional content quality evaluation directly into the crawl pipeline, preventing low-value content from entering the embedding layer at source. IGC combines distributed crawling via BullMQ and Redis, structured HTML extraction through the Mozilla Readability algorithm, a six-dimensional content quality scoring model (length, density, readability, structure, uniqueness, freshness), semantic sentence-window chunking, and a pluggable embedding pipeline supporting OpenAI, Cohere, and locally-hosted models. Embeddings are persisted in PostgreSQL with the pgvector extension, enabling approximate nearest-neighbour cosine similarity retrieval. A pilot evaluation on commodity hardware (Intel Core i3, 8 GB RAM) demonstrates stable crawl throughput of approximately 5.4 pages per second with a mean latency of 1,241 ms, while the Intelligence-Gated Crawling model filters roughly 14% of crawled pages and reduces downstream embedding noise by approximately 25%. IGC provides a scalable data-ingestion layer for retrieval-augmented generation systems, semantic search engines, and large language model dataset construction pipelines.\u003c/p\u003e","manuscriptTitle":"IGC: Intelligence-Gated Crawling for Distributed Web Content Acquisition and RAG-Ready Vectorization","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-06 09:06:54","doi":"10.21203/rs.3.rs-9039974/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ca3aecd3-2ad6-4d3e-ade2-f273a43281b6","owner":[],"postedDate":"March 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":64039620,"name":"Information Retrieval and Management"},{"id":64039621,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2026-03-06T09:06:54+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-06 09:06:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9039974","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9039974","identity":"rs-9039974","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}