← all repositories
jhljx/RecSys-Industrial-Book

A recsys textbook that ships with its own study agent

An open Chinese book on industrial recommendation systems, plus a local-first AI agent that indexes the PDFs, cites page numbers, and runs you through a 12-week course.

★502 stars Python LearningAgents
Too new to rank. Velocity needs a few days of history.
not enough history yet
star history

What it does

This is a book manuscript — 《推荐系统:工业架构与核心算法》 — covering the full pipeline of industrial recommender systems: architecture, algorithms, engineering trade-offs. The twist is that it’s packaged as an “AI-native” learning workbench: a local agent indexes the 22 published chapter PDFs and answers questions with page-level citations, while a static knowledge map (197 concept cards, 209 prerequisite links, 122 representative papers) helps you navigate the material without any JavaScript.

The interesting bit

The published PDFs are treated as the single source of truth — no LaTeX or Markdown equivalent is public — and the agent retrieves directly from them, citing PDF page numbers rather than paraphrasing from memory. There’s also a spaced-repetition-flavored course system that refuses to accept “I learned it” as evidence: mastery is verified via submitted artifacts, actual command output, and quiz scores, with review scheduled seven days later.

Key highlights

  • Local-first: no MCP, no cloud knowledge base; index and personal progress stay on your machine (git-ignored)
  • Two 12-week course tracks (new engineer, graduate) that adapt to your time budget and feedback
  • Engineering workbench commands for design contracts, code review, and evidence-tree debugging
  • Four runnable projects covering multi-path recall, two-tower version contracts, feature freshness, and overload/fallback
  • Installable Agent Skills for Claude Code and compatible clients

Caveats

  • Content is primarily Chinese; the README assumes ML/recommender fundamentals and explicitly tells beginners to go learn those first
  • License is CC BY-NC-SA — personal, non-commercial learning only; commercial reuse requires your own diligence
  • No LaTeX or Markdown source is published, so the PDFs are a hard dependency for everything

Verdict

Worth a look if you read Chinese and work on (or want to break into) industrial recommender systems — the page-cited agent and evidence-based course are a genuinely interesting publishing model. Skip it if you need English material or an actual codebase rather than a book with teaching code attached.

Frequently asked

What is jhljx/RecSys-Industrial-Book?
An open Chinese book on industrial recommendation systems, plus a local-first AI agent that indexes the PDFs, cites page numbers, and runs you through a 12-week course.
Is RecSys-Industrial-Book open source?
Yes — jhljx/RecSys-Industrial-Book is an open-source project tracked on heatdrop.
What language is RecSys-Industrial-Book written in?
jhljx/RecSys-Industrial-Book is primarily written in Python.
How popular is RecSys-Industrial-Book?
jhljx/RecSys-Industrial-Book has 502 stars on GitHub.
Where can I find RecSys-Industrial-Book?
jhljx/RecSys-Industrial-Book is on GitHub at https://github.com/jhljx/RecSys-Industrial-Book.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.