A speedrun data store, analysis engine, and racing platform.
Splits.io is how speedrunners improve through data. It gives split-by-split analysis of individual runs, viewed through a lens of all runs. On Splits.io, speedrunners share more than their time—they share their entire history of attempts, successful or not, and get feedback on how to improve long-term through statistics and comparisons with themselves and other runners in their weight class, both live (via races) and after-the-fact using historical data.
Splits.io works with LiveSplit and more than 15 other speedrunning timers. An auto-generated list can be viewed in the FAQ; new timers can self-integrate using the Splits.io Exchange Format.
Client libraries exist for the following languages. These are created and maintained by community members:
- Python:
jeremander/splitsio
- Rust:
livesplit/splits-io-api
For full API documentation and when using other languages, see the API readme.
Splits.io runs on Docker, which makes it easy and consistent to set up and run on any machine despite any unusual dependencies. The one downside is that you must first install Docker!
- Docker
- Docker Compose v2 (on macOS and Windows, this is included in the Docker install; additionally, this project has moved away from using Docker Compose v1 as it is now deprecated. More info about the v1 to v2 migration can be found on Docker Docs' Migrate to Compose V2 page)
Because of how Splits.io uses Docker, Windows requires WSL2 to be installed to run Splits.io. If you haven't done so already, follow these instructions when running Splits.io on Windows:
These steps are not required for Linux or macOS.
The first time you run Splits.io will take a while to build, so you should go grab a coffee or something:
make # shorthand for make build run
Once the output looks settled (you should see * Listening on tcp://0.0.0.0:3000
), you're good to go! Access
localhost:3000 in your browser. The first page load after a new build may also take a minute.
Note: If the page has no styling information on first run, try rebooting the server.
Now that Splits.io is running, one last step you should perform in another terminal window is:
make seed
This is required, as it sets up a client ID/secret for your local JavaScript to communicate with your local backend. It will also prepopulate your local database with some initial data.
If you have already run make
or make build
in the past, you usually won't need to rebuild everything again,
so you can instead just run the faster version:
make run
If you do need to rebuild (e.g. when adding or upgrading a dependency) but don't necessarily want a server:
make build
These steps are not required for normal operation, but you may want to perform them for specific categories of work.
Some features are built on top of links with other platforms, like Twitch
sign-in. If you want these features to work, you need to register developer
applications with the appropriate services. Copy .envrc.example
to .envrc
and follow the comments inside for details and instructions for various
platforms.
After following the instructions, run
source .envrc
make build run
to rebuild the server with your new environment variables. We recommend using direnv to automate the first step whenever you change directories!
Splits.io sends emails when users go through the "I forgot my password" flow.
In development mode, these emails are not actually sent but are instead
generated then saved to tmp/mails
.
If you want to preview a demo email in your browser, you can fiddle with the
previewers in spec/mailers/previews
then access a URL like
http://localhost:3000/rails/mailers/:mailer/:action
such as
http://localhost:3000/rails/mailers/password_reset_token_mailer/create_email.
If you're having trouble getting Splits.io running at all using the above instructions, please make a GitHub issue so we can work it out! Even if you think it's a silly issue, the fact that it's happening to you means we haven't ironed out everything (even if the only thing preventing you from setting up is better documentation!).
If you have the app up and running but are looking for insight into debugging your own changes, you can access a Rails console inside the Docker container with
make console
If you use binding.pry
anywhere in the code, once you hit the breakpoint specified use the command
make attach
in another terminal window to attach to it. To detach, make sure to exit the debug session then use the docker
attach escape sequence ctrl + p
then ctrl + q
.
If you need to attach to a container other than web
, specify a container with the syntax
make attach container=worker
To run tests from inside the Docker container, use
make test
To run only specific tests, use
make test path=spec/path/to/test/file/or/dir
We use Rubocop for code cleanliness and styling. To run it against changed files, commit your changes and run
make lint
Splits.io utilizes a few libraries for profiling our code.
Rack Mini Profiler is used to find major slowdowns in the code through the use of the badge in the top left corner of the browser window. There is also a slew of different URL parameters that you can use to get more detailed information about various aspects of the request. Details of these are explained in the readme for RMP. To get more detailed information about how code will perform in a production like environment, run
make profile
to boot the app in the profiling environment, which has most of the production flags toggled on.
DerailedBenchmarks is used to test memory over lots of requests. The commands that can be run are detailed in the readme for DB. When you have a command you want to run, use the make task like so with the options that you need
make derailed env="-e TEST_COUNT=5000 -e USE_AUTH=true" command="exec perf:mem_over_time"
The env flag is optional, so feel free to leave that blank if you have no environment variables to set.
If you change the Dockerfile or Gemfile, you'll need to run
make build
to rebuild the Docker image for your changes to apply.
If you want to reset from scratch, you can run
make clean
which will run docker compose down
, remove the bundler volume, and remove node_modules/
.
Splits.io is built in Ruby on Rails, but has some help from other pieces of infrastructure.
┌──────────────────────────────────────────────────────────┐
│AWS Application Load Balancer (splits.io) │
│┌────────────────────────────────────────────────────────┐│
││AWS Auto Scaling Group ││
││┌──────────────────────────────────────────────────────┐││
│││AWS Target Group │││
┌────────────┐ │││┌────────────────┐┌────────────────┐┌────────────────┐│││ Lambda tells Rails
│AWS RDS │ ││││AWS EC2 Instance││AWS EC2 Instance││ ││││ to parse the file
│┌──────────┐│ ││││┌──────────────┐││┌──────────────┐││ ││││ │
││PostgreSQL││◀──────┐ │││││Docker ││││Docker │││ ││││ │ ┌──────────┐
│└──────────┘│ │ │││││┌────────────┐││││┌────────────┐│││ ││││◀────│AWS Lambda│
└────────────┘ ├───┼┼┼┼┤│Rails Web ││││││Rails Web ││││ ... ││││ └──────────┘
┌───────────────┐ │ │││││└────────────┘││││└────────────┘│││ ││││ ▲
│AWS Elasticache│ │ │││││┌────────────┐││││┌────────────┐│││ ││││ │
│┌─────────────┐│ ├───┼┼┼┼┤│Rails Worker││││││Rails Worker││││ ││││ │ New file
││Redis ││◀──────┤ │││││└────────────┘││││└────────────┘│││ ││││ │ trigger
│└─────────────┘│ │ ││││└──────────────┘││└──────────────┘││ ││││ │
└───────────────┘ │ │││└────────────────┘└────────────────┘└────────────────┘│││ │
┌───────┐ │ ││└──────────────────────────────────────────────────────┘││ ┌──────┐
│AWS SES│◀──────┘ │└────────────────────────────────────────────────────────┘│ │AWS S3│
└───────┘ └──────────────────────────────────────────────────────────┘ └──────┘
│ ▲ ▲
│ │ HTTPS, WebSockets │
│ ▼ │
│ ┌────┐ │
└─────────────────────────────────────────▶│User│◀─────────────────────────────────────┘
Sends "I forgot my password" emails └────┘ File uploads/downloads (runs, race
attachments) via S3 presigned URLs
Not pictured:
- beta.splits.io, an AWS Application Load Balancer with an identical hierarchy
except pegged at 1 instance, pointing to the same external infrastructure
- livesplit-core, a Rust library with Ruby bindings that gets deployed to
containers so Rails Web can call it to parse run files
- AWS CodePipeline, which calls out to AWS CodeBuild and AWS CodeDeploy to
build and deploy code on pushes to main
Rails will synchronously parse any unparsed run before rendering it, but the asynchronous Lambda job is the preferred way for runs to be parsed because it still catches unvisited runs (e.g. in the case of a multi-file upload via drag-and-drop).
In development PostgreSQL and S3 are also Docker containers (see docker-compose.yml). Lambda is not yet implemented in development mode.
Favicons are generated by Favicon Generator and its Rails gem. To generate favicons from the source image (public/logo-imageonly.svg), run
docker compose run web rails generate favicon
Config for this generation is at config/favicon.json
.
Splits.io runs vanilla Bootstrap 4. Historically we used one of a few themes to give it some additional distinction and a professionally tailored dark mode, but we decided to switch to vanilla with our own dark mode after several bad experiences with those themes slowly falling more and more out of date from mainline Bootstrap.
If you find a security vulnerability in Splits.io, please email it privately to [email protected], as posting the vulnerability in public may allow malicious people to use it before it can be fixed. We take security matters very seriously and respond quickly to disclosures.
Splits.io uses livesplit-core for parsing runs. The parser is located in lib/parser/*
. To upgrade
it, run
make update_lsc
and commit the changes.
To generate run history charts Splits.io uses Highcharts, which requires a written license. Licensing is based on the honor system, so you do not need to enter a key anywhere. Highcharts is free to use for testing purposes.