reader/docs/getting-started.en.md

# Getting started

[Project home](../README.en.md) · [Documentation](README.md) · [简体中文](getting-started.zh-CN.md)

This guide takes you through Docker deployment, account registration, and opening your first book. To run from source, use the [development guide](development.en.md).

This guide describes the current Node implementation. CI publishes `main` (development) and short commit tags, not `latest`; availability depends on a successful build. Pin a published commit tag for repeatable deployments. Legacy Java images, `/storage` data and Legado JSON sources are not directly compatible with this version.

## 1. Prepare your environment

- Install Docker Engine or Docker Desktop and the Docker Compose plugin.
- Prepare a book directory and a separate data directory.
- Make port `8080` available, or change the host port in Compose, for example to `8090:8080`.

The container serves both the API and Web UI. No separate frontend deployment is needed. Run the following commands from the repository root.

## 2. Get the deployment configuration

```sh
git clone https://cnb.cool/hectorqin/reader.git
cd reader
```

Edit [docker-compose.yml](../docker-compose.yml) and replace the host-side book path:

```yaml
volumes:
  - /path/to/your/books:/books:ro
  - ./data:/data
```

For example, use `/srv/books:/books:ro` on Linux, or `"D:/books:/books:ro"` with Docker Desktop on Windows. The relative `./data` path is resolved from the Compose file's directory.

| Directory | Purpose |
| --- | --- |
| `/books` (`BOOKS_DIR`) | Your existing book files |
| `/data` (`DATA_DIR`) | Database, accounts, reading records, covers, upload staging, downloaded files, and plugin data |

`DATA_DIR` must be outside `BOOKS_DIR` and must not be the same directory. Startup checks this constraint. Metadata stays in the data directory; reader does not create cover files or `.calibre` metadata beside your books.

### Choose read-only or writable books

- **Browse and read only**: keep `:ro`. Add books through the host or NAS and let reader scan them.
- **Manage files in the browser**: use `:rw` and give the server process write permission. Administrators can upload, rename, move, delete, and create directories in the library file manager.

A read-only book mount still allows shelves, progress, and notes to be saved in `/data`. See [configuration](configuration.en.md) for permissions and mount details.

## 3. Start and sign in

```sh
docker compose up -d
docker compose ps
docker compose logs --tail=100 reader
```

Open `http://localhost:8080`, or `http://<server-ip>:8080` from another device.

1. Register the first account. It automatically becomes the administrator.
2. Wait for the initial scan, then open a book from the shelf or library.
3. Create accounts for other members. Public registration is disabled by default after the first account.

To allow later self-registration, set `ALLOW_REGISTRATION: "true"` in the Compose `environment` section and run `docker compose up -d` again.

The health endpoint is `/api/v1/health`. For startup failures, missing books, or write errors, see [troubleshooting](configuration.en.md#troubleshooting).

If the image is unavailable, build the checked-out source:

```sh
docker compose -f docker-compose.yml -f docker-compose.build.yml up -d --build
```

This still requires access to the Node base image and npm dependencies. Keep both `-f` arguments for subsequent management commands. For published images, set `READER_IMAGE=cnb.cool/hectorqin/reader:<published-commit-tag>` in `.env` to pin a version.

### Without Compose

On Linux/macOS with a POSIX shell, you can also run the container using absolute mount paths:

```sh
mkdir -p ./data
docker run -d --name reader --restart unless-stopped \
  -p 8080:8080 \
  -v /path/to/your/books:/books:ro \
  -v "$(pwd)/data:/data" \
  cnb.cool/hectorqin/reader:main
```

On Windows, the Compose setup above avoids shell-specific path and line-continuation syntax.

## 4. Add books and sources

### Local books

Place EPUB, TXT, PDF, CBZ/ZIP, or comic image directories in your mounted library. By default, the server runs scheduled scans every 30 minutes and checks for changes every 60 seconds. Administrators can trigger a scan manually. With a writable mount, uploads are available in the administrator file manager.

The shelf contains readable books; the library lets you browse the disk directory. Removing a book from a shelf is different from deleting its file. See the [user guide (中文)](user-guide.zh-CN.md).

### OPDS

As an administrator, add an OPDS source under Sources (书源) and configure its server URL. If authentication is required, set credentials for your current account. Open the source to browse or search its catalog and acquire books for your shelf.

### External source plugins

The generic plugin host does not parse Legado JSON. Do not enter a legacy JSON file or subscription URL as an npm package. Compatibility requires a separate plugin that explicitly supports that format; otherwise use local books or OPDS.

In Sources → Plugin management, administrators can confirm that they trust the plugin, then enter an npm package name (optionally with a version or tag) or upload a `.tgz` archive produced by `npm pack` (up to 100 MiB). Successful installations are enabled automatically; no manual deployment is needed. Add a source instance using the plugin’s source type and configure it from that instance’s management entry. One plugin can provide several independent instances. Capabilities and runtime dependencies are defined by each plugin.

Plugins run with server process privileges. Only install trusted code. The server needs Node.js and npm, plus network access for npm packages and dependencies not bundled in uploaded archives. npm lifecycle scripts are disabled, so plugins must ship prebuilt runtime files. Installation files and temporary files stay in `DATA_DIR`, never in the books directory.

See the [plugin protocol](source-plugins.md) and [extension pages](plugin-extensions.md).

## 5. Upgrade and back up

The [offline backup and restore tool (中文)](backup.zh-CN.md) checks file hashes and SQLite integrity and restores into a new directory. For older builds without this command, use a stopped-service copy as described below.

Back up both the book and data directories before upgrading. For a simple file-based SQLite backup, stop the service and copy the entire `/data` directory, including database auxiliary files, to avoid copying inconsistent live state.

```sh
docker compose stop
# Back up the host book directory and ./data here.
docker compose pull
docker compose up -d
```

Keep your data directory: it contains accounts, reading records, signing keys, and plugin state. Plugin packages are upgraded separately; disable the plugin in the UI before replacing its package.

When moving from the legacy Java version, keep a full backup and use a separate new data directory. Existing book files can be mounted, but accounts, shelves, progress, notes and source settings have no automatic migration tool. Do not copy legacy `/storage` over the new `/data`. Source-built deployments must rebuild using both Compose files instead of relying on `pull`.

## Next steps

- [User guide (中文)](user-guide.zh-CN.md): reading settings, shelves, sources, TTS, and offline access.
- [Configuration](configuration.en.md): environment variables, HTTPS, permissions, and troubleshooting.
- [Development](development.en.md): local setup, checks, Docker builds, and Android.
- [API reference (中文)](api.md): automation and integrations.