Shiori is a customized fork of the original go-shiori/shiori, shaped into a self-hosted operational reading tool. The existing Go core still handles bookmark storage, offline archiving, multiple databases and single-binary distribution, while the web interface has been rebuilt around a more practical question: what should I actually read today?
Purpose and Overview
Traditional bookmark managers are good at saving links but often leave the harder decision untouched: deciding what deserves attention next. This fork keeps Shiori bookmarks as the single source of truth and adds reading intent through ordinary tags instead of introducing a ranking model, scheduling engine or parallel queue.
The redesigned application centers on three focused spaces:
- Inbox for every active bookmark and the reading workflow.
- Archive for completed or intentionally stored material.
- Maintenance for fixing missing titles and offline snapshots in batches.
Operational Reading Model
Four tags express what a bookmark is for:
read-today— the shortlist for today.quickRead— short items that fit into small gaps.focus— deeper material that needs concentrated time.inspiration— references and ideas worth revisiting.
The interface exposes those tags as filter chips with live counts. Selecting a workflow updates the URL, while the search field narrows the visible set by title, URL or tag.
Capture, Archive and Maintenance
The + action opens a compact quick-add field: paste a URL and it goes directly into the Inbox. PDFs can also be uploaded and served back inline as bookmarks.
Archived bookmarks are separated from active reading without being removed. They can be restored or deleted from their own view.
Maintenance turns incomplete metadata into explicit work. Two batch operations identify bookmarks whose title still matches the URL and bookmarks without an offline snapshot, then refresh only the affected records.
Automation and API
The application keeps the original versioned /api/v1 surface and adds a minimal shortcut endpoint for automation clients such as iPhone Shortcuts. Optional control-header authentication supports restricted automation without changing the bookmark model.
A refresh-inbox CLI command updates title metadata and recalculates workflow tags for active bookmarks. A scheduler-friendly shell wrapper and user-level systemd timer can run that maintenance every two hours.
Architecture and Distribution
- Go backend with layered auth, account, bookmark, archiver, storage and tag domains.
- React 18 + TypeScript frontend built with Vite.
- TanStack Query for server state and React Router for focused application views.
- Tailwind CSS, Radix-style primitives and Lucide icons for the interface system.
- OpenAPI-generated client for the versioned HTTP API.
- JWT authentication with optional control-header protection and environment-driven owner bootstrapping.
- SQLite, PostgreSQL, MariaDB and MySQL storage support.
The production build is embedded into the Go application, preserving Shiori's useful deployment property: the server and web application still ship as a single binary.
Responsive Interface
The same workflow remains available on smaller screens, with navigation and filtering wrapping into a touch-friendly mobile layout rather than reducing the feature set.
Authentication
The owner account can be bootstrapped from environment variables on first launch. The rebuilt sign-in screen keeps authentication separate from the reading interface while using the same restrained visual language.
Repository
The project, setup instructions and configuration reference are available in the Shiori repository. It is distributed under the MIT license inherited from the upstream project.