Go to file
Tao Bror Bojlén b653e1a22a
move to REUSE copyright spec
2019-09-19 10:04:02 +01:00
.reuse move to REUSE copyright spec 2019-09-19 10:04:02 +01:00
.vscode add full-text search 2019-07-26 22:30:11 +00:00
LICENSES move to REUSE copyright spec 2019-09-19 10:04:02 +01:00
backend fix #97 - insular instances on graph 2019-08-31 20:24:53 +01:00
frontend release 2.8.2 2019-08-31 19:55:34 +01:00
gephi only put instances with a type on the graph 2019-08-18 17:25:24 +02:00
.dokku-monorepo fix backend app name in .dokku-monorepo 2019-07-14 13:12:35 +01:00
.gitignore frontend/cytoscape feature parity 2019-07-19 18:19:53 +00:00
.gitlab-ci.yml Lint backend 2019-08-21 12:30:47 +00:00
BILL-OF-MATERIALS.md prepare for 2.0.0 release 2019-07-20 13:01:56 +03:00
CHANGELOG.md fix #97 - insular instances on graph 2019-08-31 20:24:53 +01:00
README.md update ES index on crawl 2019-08-02 19:03:21 +03:00
docker-compose.yml add full-text search 2019-07-26 22:30:11 +00:00
example.env refactor/elixir backend 2019-07-14 11:47:06 +00:00
netlify.toml fix direct links on netlify 2019-07-21 10:23:50 +03:00
nlnet-logo.png add nlnet logo to about page 2019-07-27 16:55:00 +03:00
screenshot.png add deployment instructions to README 2019-07-20 13:45:34 +03:00


fediverse.space 🌐

The map of the fediverse that you always wanted.

Read the latest updates on Mastodon: @fediversespace

A screenshot of fediverse.space

  1. Requirements
  2. Running it
  3. Commands
  4. Privacy
  5. Deployment
  6. Acknowledgements


Though dockerized, backend development is easiest if you have the following installed.

  • For the scraper + API:
    • Elixir
    • Postgres
  • For laying out the graph:
    • Java
  • For the frontend:
    • Node.js
    • Yarn

Running it


  • cp example.env .env and modify environment variables as required
  • docker-compose build
  • docker-compose up -d phoenix
    • if you don't specify phoenix, it'll also start gephi which should only be run as a regular one-off job
  • Create the elasticsearch index:
    • iex -S mix app.start
    • Elasticsearch.Index.hot_swap(Backend.Elasticsearch.Cluster, :instances)


  • cd frontend && yarn install
  • yarn start



./gradlew shadowJar compiles the graph layout program. java -Xmx1g -jar build/libs/graphBuilder.jar runs it. If running in docker, this means you run

  • docker-compose build gephi
  • docker-compose run gephi java -Xmx1g -jar build/libs/graphBuilder.jar lays out the graph


  • yarn build creates an optimized build for deployment


This project doesn't crawl personal instances: the goal is to understand communities, not individuals. The threshold for what makes an instance "personal" is defined in the backend config and the graph builder SQL.


You don't have to follow these instructions, but it's one way to set up a continuous deployment pipeline. The following are for the backend; the frontend is just a static HTML/JS site that can be deployed anywhere.

  1. Install Dokku on your web server.
  2. Install dokku-postgres, dokku-monorepo, dokku-elasticsearch, and dokku-letsencrypt.
  3. Create the apps
  • dokku apps:create phoenix
  • dokku apps:create gephi
  1. Create the backing database
  • dokku postgres:create fediversedb
  • dokku postgres:link fediversedb phoenix
  • dokku postgres:link fediversedb gephi
  1. Set up ElasticSearch
  • dokku elasticsearch:create fediverse
  • dokku elasticsearch:link fediverse phoenix
  1. Update the backend configuration. In particular, change the user_agent in config.exs to something descriptive.
  2. Push the apps, e.g. git push dokku@<DOMAIN>:phoenix (note that the first push cannot be from the CD pipeline).
  3. Set up SSL for the Phoenix app
  • dokku letsencrypt phoenix
  • dokku letsencrypt:cron-job --add
  1. Set up a cron job for the graph layout (use the dokku user). E.g.
0 2 * * * /usr/bin/dokku run gephi java -Xmx1g -jar build/libs/graphBuilder.jar
  1. (Optional) Set up caching with something like dokku-nginx-cache

Before the app starts running, make sure that the Elasticsearch index exists -- otherwise it'll create one called instances, which should be the name of the alias. Then it won't be able to hot swap if you reindex in the future.


NLnet logo

Many thanks to NLnet for their support and guidance of this project.