{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreidglsicwvyacxfbpxiihqp4wd7arcsldpdihndusalx6y3vtibhfq",
"uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mfd5wdokofa2"
},
"path": "/t/best-ai-architecture-for-processing-and-querying-large-pdfs-7000-pages-with-fast-response-time/1374726#post_1",
"publishedAt": "2026-02-20T21:41:25.000Z",
"site": "https://community.openai.com",
"textContent": "Hi everyone,\n\nI’m building an AI-powered educational assistant that must answer questions grounded in very large PDFs (up to ~4000 pages per document). I’m currently using a RAG-based setup, but I’m facing serious production issues and would appreciate architectural guidance.\n\n### Current Problems:\n\n 1. **Very high latency**\n\n * Responses take 25+ seconds.\n\n * Sometimes even longer with complex queries.\n\n 2. **Missing information in responses**\n\n * The system retrieves only partial sections.\n\n * Important parts of the document are ignored.\n\n * Answers feel incomplete or fragmented.\n\n 3. Requirements:\n\n\n\n * Fast response time (<3 seconds ideally)\n\n * High-quality, well-structured answers\n\n * Accurate grounding with page references\n\n * Ability to handle 4000+ pages reliably\n\n * Production-ready and scalable\n\n\n\n\nAny one have a better Ai pipeline, or knew how to implement it in a proper way ?",
"title": "Best AI Architecture for Processing and Querying Large PDFs (7000 Pages) with Fast Response Time"
}