An app that makes it possible to schedule posts on Lemmy.

Am I the only one who thinks it's funny how the name "Lemmy Schedule" sounds like "let me schedule"?

• Lemmy Schedule
  • Interesting parts
  • How it works?
  • Building the app locally
    • Prerequisites
    • Configuring
      • Job transports
      • File uploading
      • Cache
    • Building
    • Running
  • Self-hosting - AWS
    • Prerequisites:
    • Environment variables
    • Deploying
  • Self-hosting - docker
    • Configuration
    • Volumes
    • Ports
    • Running
      • Docker compose

• AppAuthenticator - the authentication process, here you can make sure that I'm not a villain trying to steal your precious credentials and then use the power of impersonating you to take over the world.
• CreatePostJob - the job that gets saved to schedule your post, as you can see I store your JWT to impersonate you, but sadly it's impossible to do without.
• EventBridgeTransport - here you can see that the job is deleted immediately after the job is triggered (see the line that reads 'ActionAfterCompletion' => ActionAfterCompletion::DELETE)

It uses the excellent Symfony framework and even more excellent Symfony Messenger.

Basically, it creates a Messenger job and configures a delay for the job that is the same as the time of the post. The job is then received, processed and deleted (depends on your particular messenger backend).

The default deployment using the provided serverless configuration has a little twist on that: to be fully serverless, the messenger backend is faked to be an EventBridge service and it only allows creating jobs, which it does by creating a scheduled job which in turn posts to a console command app:run-sync.

The console command takes the job, overwrites the transport config to force it to run synchronously in the command and thus the need for a queue consumer is avoided. I'm still not sure whether this is a genius idea or a blasphemy. But it saves me money for hosting, so I'm gonna go with genius.

Instead of database it uses a cache backend which can be any PSR-6 backend, the default configuration uses a DynamoDB one.

Note that this is a tutorial for a fully local deployment, depending on the method you'll be using for deployment (Docker, AWS serverless), you might not need to set up everything mentioned below, feel free to skip to the specific guides below and use this only as a reference if you want a deeper understanding of how the app works.

Here's a brief tutorial on how to build a production version of the app.

• php (8.2, intl and json extensions), composer etc.
• serverless framework
• yarn (if you use a different node package manager, you can adapt the commands to it)

All configuration is made using environment variables, those can either be real environment variables or ones set in .env.local file.

If an option is marked as required, but also has a default value, you don't have to provide your own value if you're ok with the default

Job transport is what manages where the scheduled jobs are stored and where does the messenger component consume them from. You can see the full list of officially supported transports in the official Symfony documentation.

If you're not familiar with the Messenger component, I advise you to go with Redis which is the simplest to configure. The Redis DSN looks like this: redis://localhost:6379/messages.

This configures the server (localhost), port (6379) and key (messages) where the jobs will be stored. For more complex examples visit the official documentation.

Currently, there are two options for storing uploaded files — locally or in AWS S3. In either case, the files are deleted once the job that references them is triggered.

These correspond to these FILE_UPLOADER_CLASS values:

Note that this environment variable needs to be set prior to compiling the service container

• local - App\FileUploader\LocalFileUploader
• S3 - App\FileUploader\S3FileUploader

For local, you need to also specify LOCAL_FILE_UPLOADER_PATH which is a directory in which the file uploads will be stored. You can use a special variable %kernel.project_dir% that gets replaced with the path to where the project is stored.

If you go with S3, you also need to specify S3_FILE_UPLOADER_BUCKET which is the name of the bucket where the files will be stored.

File providers are handlers for uploading images. Some of them might need configuration before they show up in the UI.

Currently supported ones:

• instance upload - the default, always supported
• Imgur - an Imgur access token is required

If you're not using DynamoDB for cache, delete the file config/packages/prod/rikudou_dynamo_db_cache.yaml (before compiling the service container).

• Install the production php dependencies - composer install --no-dev --no-scripts
• Install node dev dependencies - yarn install
• Build static JS assets - yarn build
• Compile the production service container - ./bin/console cache:warmup --env=prod

Done!

If you want to test it out, run the development php server using php -S 127.0.0.1:8000, otherwise follow some guide to set up a webserver, Apache being the easiest solution.

If you intend to self-host using docker, skip this section

Note: This flow can also always be checked in publish.yaml

• read the section on building the app
• aws cli
• a domain in AWS Route53

These variables are provided automatically and you cannot change them:

• APP_ENV
• APP_SECRET
• DYNAMODB_CACHE_TABLE
• MESSENGER_TRANSPORT_DSN
• S3_FILE_UPLOADER_BUCKET
• FILE_UPLOADER_CLASS
• ROLE_ARN
• APP_CACHE_DIR
• APP_LOG_DIR

You need to set these environment variables (as real environment variables, not as part of .env.local):

• DOMAIN_NAME
• DOMAIN_ZONE - the zone ID of the Route53 domain

You may set these environment variables as well (as real environment variables, not as part of .env.local):

• SINGLE_INSTANCE_MODE
• IMGUR_ACCESS_TOKEN
• UNREAD_POSTS_BOT_JWT
• UNREAD_POSTS_BOT_INSTANCE
• FLAG_COMMUNITY_GROUPS

• Follow the guide on building
• Deploy using the serverless command: serverless deploy --stage prod --verbose
  • The --stage prod can be changed to --stage dev for a dev build
  • The --verbose flag can be skipped if you want less output

If everything works out, you should be able to visit the domain which you specified in DOMAIN_NAME and the app should be running there!

You need to configure these variables:

• APP_SECRET - a random string around 32 characters length
  • you can use, for example, this command to generate one: hexdump -vn16 -e'4/4 "%08x" 1 "\n"' /dev/urandom

Other variables which might need changing:

• DEFAULT_INSTANCE - set this to your preferred instance, otherwise the default lemmings.world will be used
• MESSENGER_TRANSPORT_DSN - the transport name is by default set to expect a container with hostname redis, you might need to change this if your setup is different
  • Read the documentation on the transport in this README
• APP_ENV - you might want to change this to dev if you're debugging the app, otherwise leave it out (or set to prod)
• FILE_UPLOADER_CLASS - you may want to change how uploaded files are handled
  • Read the documentation on uploading files in this README
• SINGLE_INSTANCE_MODE - set to 1 if you want to run in single instance mode where only people from instance set in DEFAULT_INSTANCE can log in
• IMGUR_ACCESS_TOKEN - the access token to use with Imgur, leave empty to not support Imgur
• UNREAD_POSTS_BOT_JWT - JWT token of the bot user that will send unread post reports
• UNREAD_POSTS_BOT_INSTANCE - the instance of the aforementioned bot user
• FLAG_COMMUNITY_GROUPS - whether to enable community groups, an experimental feature

Some permanent files are accessible at these locations:

• /opt/runtime-cache - used for storing configuration etc., MUST be bound to a volume
• /opt/uploaded-files - directory for uploaded images, MUST be bound to a volume

The app runs on port 80 inside the container, bind the port to any port you want.

Taking the above into account, here is a docker command used for the production build (replace the APP_SECRET):

• docker run -d -it -p 8000:80 -v files:/opt/uploaded-files -v cache:/opt/runtime-cache -e DEFAULT_INSTANCE=my.instance.tld -e APP_SECRET=5d640026bb762a0e874c1e1f2656e079 ghcr.io/rikudousage/lemmy-schedule:latest

If you have a Redis instance under different URL (replace redis.example.com with your hostname and lemmy_schedule with the Redis key you want to use):

• docker run -d -it -p 8000:80 -v files:/opt/uploaded-files -v cache:/opt/runtime-cache -e DEFAULT_INSTANCE=my.instance.tld -e APP_SECRET=5d640026bb762a0e874c1e1f2656e079 -e MESSENGER_TRANSPORT_DSN=redis://redis.example.com/lemmy_schedule ghcr.io/rikudousage/lemmy-schedule:latest

You can easily run the app using docker compose like so:

version: "3.7"

services:
  redis:
    image: redis
    hostname: redis
    command: redis-server --save 60 1 --loglevel warning # make Redis dump the contents to disk and restore them on start
    volumes:
      - redis_data:/data
  lemmy_schedule:
    image: ghcr.io/rikudousage/lemmy-schedule:latest
    ports:
      - "8000:80" # replace 8000 with the port you want your app to run on
    environment:
      APP_SECRET: $APP_SECRET # actually create the secret, don't just use this value
      DEFAULT_INSTANCE: my.instance.tld
    volumes:
      - ./volumes/lemmy-schedule-cache:/opt/runtime-cache
      - ./volumes/lemmy-schedule-uploads:/opt/uploaded-files
    depends_on:
      - redis

volumes:
  redis_data: