The EtherBranch knowledge commons

EtherBranch is more than a bookstorefront. It is a closed loop: find a text, purchase it, own it on your Library shelf, study it in Lab with an AI tutor that only knows that book, and eventually sit a proctored Exam. The machine-learning pieces are not decorative—they decide what you review next, what the tutor cites, and how mastery decays over time.

This wiki is written for students, instructors, and engineers who want the full story. Product pages stay short; here we take the scenic route through retrieval, caching, Bloom's taxonomy, and SM-2 spaced repetition as they are implemented (and stubbed) in the current codebase.

1. Discover on the storefront or Library.
2. Checkout on checkout.etherbranch.com—payment is captured and owned titles are written to BookOwned.
3. Journal becomes your hub: Lab and Exam are first-class destinations.
4. Lab opens a book-scoped AI session. RAG retrieves the right passages; CAG keeps hot curriculum context warm; SM-2 schedules what you forgot.
5. Exam uses Zoom-backed proctoring when you are ready to demonstrate mastery.

Production surfaces are intentionally separated so each can evolve its own density of UI:

• www.etherbranch.com — bookstore & discovery
• library.etherbranch.com — immersive stacks & ownership
• checkout.etherbranch.com — shipping + payment
• journal.etherbranch.com — post-login hub
• lab.etherbranch.com — AI Study Lab
• exam.etherbranch.com — Zoom proctoring

Locally, the same routes live under one Next.js app (/library, /lab, …) with middleware rewriting subdomain hosts.

Until the GPU and production LLM are wired, Lab streams from lib/lab/mockInference.ts. Every export is marked TODO(AI-INTEGRATION). The RAG and CAG designs below describe the target architecture the mock already mimics—so swapping the backend is a clean cut, not a rewrite of the product.