# reminder-rs

Reminder Bot for Discord.

## How do I use it?

I offer a hosted version of the bot. You can invite it with: **https://invite.reminder-bot.com**.
The catch is that repeating
reminders are paid on the hosted version of the bot. Keep reading if you want to host it yourself.

You'll need rustc and cargo for compilation. To run, you'll need Python 3 still (due to no suitable
replacement for dateparser in Rust)

### Build APT package

Recommended method.

By default, this builds targeting Ubuntu 20.04. Modify the Containerfile if you wish to target a
different platform. These instructions are written using `podman`, but `docker` should work too.

1. Install container software and Make: `sudo apt install podman make`.
2. Install database server: `sudo apt install mysql-server-8.0`. Create a database called
   `reminders`
3. Install SQLx CLI matching the project's SQLx version: `cargo install sqlx-cli --version 0.8.2 --locked`.
4. Install the local build requirements (Rust, Node.js/npm, GCC, CMake, pkg-config, and OpenSSL
   development headers), then install dashboard dependencies: `npm ci --prefix reminder-dashboard`.
5. Export the URL of your build/test database:
   `export DATABASE_URL="mysql://user[:password]@localhost/reminders"`.
6. From the source code directory, execute `sqlx migrate run`.
7. Run `make` (or `make build`). This prepares SQLx query metadata in `.sqlx`, runs the Rust tests,
   builds the `reminder-rs` container image, and runs `cargo deb` inside it. The resulting package is
   written to `target/debian/`.

Preparation and database-dependent tests use the local database. The container packaging step uses
the prepared `.sqlx` metadata with `SQLX_OFFLINE=true`, so it does not need database access, including
when a `.env` file is present in the mounted source directory.

The stages can also be run separately: `make prepare`, `make test` (includes preparation), and
`make image` (includes preparation and tests). Override `IMAGE` to change the container image tag,
or `PODMAN` to use another container command, for example `make PODMAN=docker`.

### Compiling for other target

1. Install requirements:
   `sudo apt install gcc gcc-multilib cmake libssl-dev build-essential python3-dateparser`
2. Install rustup from https://rustup.rs
3. Install the nightly toolchain: `rustup toolchain default nightly`
4. Install database server: `sudo apt install mysql-server-8.0`. Create a database called
   `reminders`.
5. Install `sqlx-cli`: `cargo install sqlx-cli`.
6. Run migrations: `sqlx migrate run`.
7. Set environment variables:
    * `DATABASE_URL` - the URL of your MySQL database (`mysql://user[:password]@domain/database`)
8. Build: `cargo build --release`

### Configuring

Reminder Bot reads a number of environment variables. Some are essential, and others have hardcoded
fallbacks. Environment variables can be loaded from a .env file in the working directory.

__Required Variables__

* `DATABASE_URL` - the URL of your MySQL database (`mysql://user[:password]@domain/database`)
* `DISCORD_TOKEN` - your application's bot user's authorization token

__Other Variables__

* `MIN_INTERVAL` - default `600`, defines the shortest interval the bot should accept
* `LOCAL_TIMEZONE` - default `UTC`, necessary for calculations in the natural language processor
* `SUBSCRIPTION_ROLES` - default `None`, accepts a list of Discord role IDs that are given to
  subscribed users
* `CNC_GUILD` - default `None`, accepts a single Discord guild ID for the server that the
  subscription roles belong to
* `PYTHON_LOCATION` - default `/usr/bin/python3`. Can be changed if your Python executable is
  located somewhere else
* `THEME_COLOR` - default `8fb677`. Specifies the hex value of the color to use on info message
  embeds

### Optional dashboard image uploads

Image URL inputs offer an **Upload image** button when all of these settings are supplied:

* `S3_BUCKET` — destination bucket name.
* `S3_REGION` — bucket region, for example `eu-west-2`.
* `S3_ACCESS_KEY_ID` and `S3_SECRET_ACCESS_KEY` — credentials with `s3:PutObject` and
  `s3:DeleteObject` permissions for the bucket's `uploads/` prefix.
* `S3_PUBLIC_URL` — public base URL for the bucket or CDN, for example
  `https://my-bucket.s3.eu-west-2.amazonaws.com`. Objects must be publicly readable through
  this URL; the application does not change bucket policies or set object ACLs.

Optionally set `S3_ENDPOINT` for an S3-compatible provider; this uses path-style addressing.
The ONNX model in `assets/nsfw/model.onnx` is embedded in the binary at compile time using
`include_bytes!`; no model file, download or model-path setting is needed at runtime, including
in containers. The bundled model adds approximately 16.5 MiB to the binary and is initialized
once at startup only when S3 uploads are configured and NSFW checks are enabled.
See [model provenance](assets/nsfw/README.md).
Missing or invalid S3 configuration disables uploads while the dashboard continues to accept
image URLs. If the bundled model fails to initialize, uploads are disabled as well.

`NSFW_CHECK_ENABLED` defaults to `true`. Set it to `false`, `0` or `off` to skip NSFW
moderation and model initialization; `true`, `1` or `on` enables checks. Values are
case-insensitive. Empty or unrecognized values keep checks enabled. Restart the application
after changing this setting. Image format validation, size limits and uploader tracking still
apply when moderation is disabled.

When enabled, `POST /dashboard/api/images` accepts the image file as the raw request body with
`Content-Type: application/octet-stream`, using an authenticated dashboard session. It accepts
PNG, JPEG, GIF and WebP up to 8 MiB and 16 million pixels. Files are decoded, flattened onto a
white background and saved as static PNGs; animation and metadata are removed. By default, the `nsfw` crate
checks the exact pixels that will be published, rejecting images when the combined Porn, Hentai
and Sexy probability is at least 50%. Rejections return HTTP 451 and the dashboard shows an
information modal. Moderation failures reject the upload without publishing it.

Successful uploads return `{ "url": "https://.../uploads/....png" }` and record the uploader's
Discord ID, public URL and upload time in `image_uploads`. The database migration runs with the
other migrations. If recording fails, the application attempts to remove the S3 object and
returns an error. The endpoint is not mounted when uploads are disabled.

For local image-based moderation tests, place split image pairs in `tests/fixtures/nsfw/`
and `tests/fixtures/non_nsfw/`. The [fixture instructions](tests/fixtures/README.md) describe
the splitter script and the opt-in tests that reconstruct the images before classification.

## Running with Docker

A `compose.yml` file is provided to aid in running the bot agnostically using docker.

* Populate a `.env` file as in `conf/default.env`
* Add the additional variable `ROCKET_SECRET_KEY` with a key generated from
  `head -c64 /dev/urandom | base64`
* Run `docker compose up`

Please note that this is _not_ production-ready when run via compose. We do not offer a way for
backing up of your data, or a way to run the dashboard securely via HTTPS, which is required for
OAuth.