BookNest — A Personal Library with Book Knowledge Maps

BookNest is a personal library for books and magazines that runs on your own machine, using SQLite and covers stored in a local folder — no account, cloud service or network connection needed. Besides reading status, priority, notes and tags, each book can have a “knowledge workspace”: a booknest-knowledge/v1 JSON document that an AI researches from the app’s prompt and you import to be validated. It holds the real table of contents (kept apart from the AI-synthesised content map), terms, relations, diagrams, reading paths and related books. Every statement is labelled verified or synthesised and cites its source, and chosen books can be published as a read-only page. A Chrome extension imports books from an Amazon Store.

1. 01Many technical books go half-read or unopened; with only a title and a cover you can’t tell what a book is about, who it is for, or where to start.
2. 02An AI can summarise a book but mixes fact with interpretation, and showing an AI-synthesised map as the real table of contents would mislead the reader.
3. 03A library is personal data that shouldn’t have to live with an outside service, and it must be easy to back up and move.
4. 04Typing book details one at a time is slow, and many online shops offer no way to export a list.

1. 01A local-first library: Next.js + Prisma + SQLite (11 models), covers on local disk, no account or cloud. A first run creates the database from the committed migration, and Settings can build a ZIP backup of the database and covers.
2. 02Books and magazines in one place: several authors, ISBN-10/13 with check-digit validation, edition, year, publisher, tags, status (Want to read, Reading, Finished, On hold, Archived), priority and favourites, duplicate detection before saving, import from a book-page link (with Cheerio) or pasting a cover with Ctrl/⌘+V. Magazine issues track individual articles and whether you read them.
3. 03The booknest-knowledge/v1 contract is validated in layers with Zod: sources, overview, technical profile, table of contents (6 entry kinds), content map, terms, relations (6 types), diagrams (7 types), reading paths and errata. The key rules: a real contents entry needs a source, a statement labelled “verified” needs a citation, and the JSON must match the identity of the book being edited (title, authors, edition, year, ISBN).
4. 04The loop with an AI: a button copies a prompt pre-filled with that book’s title, authors, edition, year and ISBN, with a JSON template showing every collection. You import what the AI returns, the editor validates it immediately and points at the offending location, and a revision number stops a save overwriting an edit made in another tab.
5. 05A tabbed reader for the knowledge (overview, real contents, content map, terms, relations, diagrams, reading paths, sources), whole-book search with ⌘K, and related books found through shared topics.
6. 06The relations graph and diagrams use React Flow, with the layout computed by ELK in a Web Worker so the page doesn’t freeze. It handles cycles, duplicate edges and self-loops, exports SVG or PNG, and opens full screen.
7. 07Publish only what you choose: enable “publish this book’s data” for a read-only /share/books/[id] page, run npm run publish to export standalone HTML for the published books (without personal notes, reading status, priority, favourites or edit controls), and npm run make for an offline book list with a Load more button.
8. 08A Chrome extension, “Amazon Store Importer” (Manifest V3): it reads the product cards rendered in the open tab, clicks Load more for you, verifies each item against Amazon’s metadata, then sends only the Store URL, ASIN, title, author, publication year and cover to the local BookNest. /books/store-import marks books already owned. The extension doesn’t read passwords, payment details or browsing history.

1. 01A library where opening a book tells you what it covers, who it suits, what to know first and which chapters to read first — with what the book says clearly separated from what an AI synthesised.
2. 02All the data stays on the machine: one SQLite file plus a covers folder, easy to back up and move as a ZIP.
3. 03Three verification scripts: verify:knowledge (the JSON validator, books with and without Parts, revisions, persistence, deletion and public export against a temporary database), verify:graphs (diagram layout) and verify:graphs:browser (diagrams checked in a real browser with Playwright).

• A Next.js 15 + Prisma + SQLite web app (11 models) with migrations, ZIP backup and a Settings page
• A book and magazine library with filtering, search, duplicate detection, link import, and author, publisher and tag management
• The booknest-knowledge/v1 validator, the AI prompt, a JSON template, and the knowledge editor and reader
• Relations graph and diagrams with React Flow + ELK (Web Worker) and SVG/PNG export
• A read-only share page, standalone HTML export (npm run publish) and an offline book list (npm run make)
• The Chrome extension Amazon Store Importer (Manifest V3, version 1.3.0)
• The verify:knowledge, verify:graphs and verify:graphs:browser scripts

• Next.js
• React
• TypeScript
• Tailwind CSS
• Prisma
• SQLite
• Zod
• React Flow
• ELK
• Cheerio
• Chrome Extension (Manifest V3)
• Playwright

• 2026Catadex — A Directory of Specialist Catalogs1,248 specialist data catalogs in 34 fields, explored through cards, a relationship graph and guided trails, and searchable in Thai and EnglishPersonal
• 2026Atlas — Personal Research WorkspaceCapture sources, connect topics, test claims against evidence and write knowledge you can revisit — a private web app I designed end to end, from the schema to the security modelPersonal
• 2026graph.phatphum.me — Visual Knowledge AtlasOne knowledge model, several ways to read it: a dependency graph, a roadmap, an outline and a focus view — with Frontend, Computer Networks and Cybersecurity curricula I wrote and validated myselfPersonal