Dev Publishes a Build-It-From-Documents-Up RAG Book, Everything Runs Locally
2redditornot · reddit · 2026-08-19
The author released Volume I of a RAG book on Leanpub, written as a build-along story rather than a reference, with every line of code executed locally on a laptop. Volume I covers the offline half:
- Extraction across PDF, Word, xlsx, HTML, email, plus OCR for a 1994 scan with no text layer;
- Cleaning with receipts for everything removed;
- Chunking (fixed-size, recursive, structure-aware) and why fixed-size is wrong for real documents;
- Embeddings, similarity, model choice, and a vector database.
He invites pushback on two claims: exact brute-force search is fine into the hundreds of thousands of vectors (early ANN indexes are "a forklift for a bookshelf"), and using all-MiniLM-L6-v2 throughout the pilot. Volume II will cover generation, citations and evaluation. Chapter 1 is free.
More from coding & agent
- Upcoming Agent course series: First video on evaluating evals — ben_burtenshaw · 2026-08-19
- Claude Code plugin rewrites output to plain English using local LLMs — tom_doerr · 2026-08-19
- Users need free open-source CI and seamless mirroring to switch from GitHub — JoshuaJBouw · 2026-08-19
- Agent Evaluation Dishonesty: Research Reveals Router and Reward Blind Spots — Brilliant-Tour6466 · 2026-08-19
- Developer argues most LLM agent projects are noise, prompts suffice — JoshuaJBouw · 2026-08-19
- Microsoft's Agent Lightning boosts SWE-bench score to 56.4% with 6K samples — omarsar0 · 2026-08-19