An open-source, native AI reading workspace for macOS.
Reader keeps the book, your notes, web research, and an AI conversation in one quiet workspace. Import PDF, EPUB, and TXT files; read in a native interface; highlight and annotate passages; simplify difficult writing; ask questions about a selection, the current page, everything up to your position, or the whole book; and open chat, a dictionary, or the web beside the text.
The library and reading experience are local-first. Books, progress, highlights, notes, search, and saved conversations remain usable without an account or backend. Supabase sync and AI features are optional additions.
Features
- Native reading: import and read PDF, EPUB, and plain-text books in a focused macOS interface built with SwiftUI, PDFKit, and WebKit.
- Contextual AI: chat about a selected passage, the current page, the book up to your reading position, or the complete book.
- Inline understanding: rewrite a difficult passage in clearer language without leaving the page.
- Highlights and notes: preserve selections, annotations, and reading context alongside the book.
- One reading workspace: place the book beside AI chat, a system dictionary, another book, or a native browser.
- Local library: organize books into folders, search across the library, resume progress, and keep local data in SQLite.
- Optional private cloud: sync originals, reading state, annotations, and searchable book context through your own Supabase and Vercel backend.
- Local-first failure model: opening, reading, searching, highlighting, and annotating do not depend on cloud availability.
This monorepo contains the native Reader app and its optional Hono, Supabase, and Vercel backend.
Run
Requires macOS 15 or later and Xcode 16 or later. Open Reader.xcodeproj
in Xcode and run the Reader scheme. Node.js and backend credentials are not
required to build the native app. Cloud and AI features need a backend. The
checked-in backend URL is the local Vercel development address at
http://127.0.0.1:3000; it never sends a fork to maintainer infrastructure.
Follow SETUP.md to create your own Supabase and Vercel projects and
point the app at them.
From Terminal:
DEVELOPER_DIR=/Applications/Xcode.app/Contents/Developer \ xcodebuild \ -project Reader.xcodeproj \ -scheme Reader \ -destination 'platform=macOS' \ CODE_SIGNING_ALLOWED=NO \ build test
The project is generated with XcodeGen:
The generated .xcodeproj is committed so XcodeGen is helpful, not required, for someone opening the app.
Install and check the backend from the same repository root:
make setup make backend-check
Run the backend locally through Vercel:
make backend-link make backend-dev
Deploy a preview from the same root:
Run both native and backend checks:
The backend package has more detail in Backend/README.md.
The complete app-only and cloud setup is in SETUP.md.
I know you're just having your agent run this, so here's a guide made specifically for it lol: AGENT_SETUP.md.
Repository layout
Reader/ Native Swift and SwiftUI source
ReaderTests/ Native tests
Backend/ Hono, Supabase, and Vercel backend
TestCorpus/ Reproducible public-domain format corpus
How it works
The app owns a shared Reader Core for search, locators, selections, annotations, progress intent, and capabilities. Format adapters translate those concepts into the rendering primitive that fits:
SwiftUI reading experience
└── Reader Core
├── PDF adapter → PDFKit
└── Reflowable adapter → owned EPUB/TXT parser + hardened WKWebView
Reader uses ZIPFoundation for archive extraction and GRDB for SQLite persistence. Rendering remains owned by Reader through PDFKit and its EPUB/TXT WebKit adapter.
Mac first, never Mac only
Reader is built natively in Swift, making it easy to bring to iOS and iPadOS; that goal guides technical decisions so one shared codebase can support macOS, iOS, and iPadOS.
What is implemented
- Native macOS app with SwiftUI
- Local PDF, EPUB, and TXT import into an app-managed copy
- PDFKit fixed-page rendering
- Custom EPUB/TXT extraction and WebKit reading surface
- Shared native search and annotation model above both adapters
- Embedded EPUB covers and first-page PDF covers
- Supabase account entry plus optional, private cloud copies of imported originals
- Import-time background preparation for whole-book AI search and annotations
- Authenticated launch and foreground reconciliation for existing or interrupted local books
- Truthful cloud status derived from confirmed file and searchable-text preparation
- Local SQLite persistence for library entries, progress, highlights, and conversations
- EPUB is vertically reflowed by section; pagination and full standards coverage are still open
- Local reading remains available when cloud sync or AI preparation fails
Contributing
See CONTRIBUTING.md for the development workflow and contribution terms.
License
Reader is open source under the MIT License. You may use, copy, modify, publish, distribute, sublicense, and sell copies of the software subject to the license terms.