This is an example application based on GitHub that's meant to model GitHub's
permissions system. The app is implemented as multiple backend services (in the
services/
directory) that use Oso Cloud as a shared central authorization
system and a React frontend (in the frontend/
directory).
This application is built as an example for Oso Cloud.
This is an example application based on GitHub that's meant to model GitHub's
permissions system. The app uses the oso-cloud
Python
library to model, manage, and enforce authorization.
The Oso Cloud documentation is a good reference for more information on Oso's Python library.
This is an example application based on GitHub Actions that's meant to model
GitHub Actions's permissions system. The app uses the
oso-cloud
Node.js library to model, manage, and enforce
authorization.
The Oso Cloud documentation is a good reference for more information on Oso's Node.js library.
Install the oso-cloud
CLI by following the instructions in the
Oso Cloud Quickstart
You can optionally set up your environment for local testing by running make setup
from the root directory of the repository. This will do the following:
- Install the appropriate version of the Oso Cloud local development binary for your system.
- Copy the pre-commit hook from the
scripts
directory to.git/pre-commit
if one doesn't already exist there.- This will automatically validate the syntax of
policy/authorization.polar
before committing it.
- This will automatically validate the syntax of
If you configured your environment for local testing above, then you can run the policy tests locally by running the following command from the root directory of the repository:
OSO_AUTH=e_0123456789_12345_osotesttoken01xiIn make test-policy
The OSO_AUTH
variable is set to the local development API key as documented in the
Setup section of the Local Development Guide.
It will only work against the local development binary.
This will:
- start the local development binary
- Run the policy tests in
policy/authorization.polar
against the local development binary
NOTE: running the test suite against an Oso Cloud instance will reset all data in that instance. This ensures each test starts from a clean slate.
To run the test suite, which, again, will reset data and perform many
authorization requests against the target Oso Cloud instance, grab your API
key from https://ui.osohq.com/dashboard and export it as the OSO_AUTH
environment variable.
$ export OSO_AUTH="0123456789"
$ make test
The first time you run these services, you'll need to upload your policy:
make update-policy
You only need to do this once. (If you make changes to the policy, you'll need to run it again.)
Run GitClub in one terminal:
make -C services/gitclub
Run the Jobs Service in another terminal:
make -C services/jobs
You can run all the services at once, along with a Oso Cloud local development binary and Postgres instance, using Docker Compose:
docker-compose up
This repo's docker-compose.yml
file starts the following services:
- GitClub on port 5000
- Jobs Service on port 5001
- Oso Cloud local development binary on port 8081
- Postgres on port 5432
- Next.js frontend on port 8000
If docker-compose up
ran successfully, you should see the following output.
frontend-1 | yarn run v1.22.19
frontend-1 | $ next start
frontend-1 | ready - started server on 0.0.0.0:8000, url: http://localhost:8000
If this is your first time running docker-compose up
, you will also need to seed your local Postgres database by running make seed
.
If you run into the following error when running docker-compose up
, run ALTER TABLE users DROP CONSTRAINT IF EXISTS pg_type_typname_nsp_index;
to clean up the dangling unique constraint.
psycopg2.errors.UniqueViolation: duplicate key value violates unique constraint "pg_type_typname_nsp_index"
You can run this code from Python using the following:
with engine.connect() as conn:
conn.execute('ALTER TABLE users DROP CONSTRAINT IF EXISTS pg_type_typname_nsp_index')
$ cd frontend
$ yarn
$ yarn start
- TypeScript / React / Reach Router
The GitClub service runs on port 5000; the Jobs service runs on port 5001. The frontend knows to make requests to 5001 for Jobs and 5000 for everything else. In a real production scenario, these disparate backend APIs would probably be abstracted behind an API gateway or similar.
The GitClub service uses cookies to manage sessions. The Jobs service looks
for a special x-user-id
header that contains the logged-in user's super secret
token... which is also their ID... which is also their email. In a production
scenario, the authentication system should be a lot more secure. It doesn't
really matter for our purposes, where we really only care about showing off
authorization might.
If you want to be able to debug/test the backend without running the frontend and logging in, you can use the following to save a session locally:
curl -c gitclub.cookies -H "Content-Type: application/json" -X POST -d '{"username": "john@beatles.com"}' localhost:5000/session
curl -b gitclub.cookies localhost:5000/orgs/1
Try disabling Airplay, Airplay can intercept requests to localhost:5000
.
In the Chrome console, check the "server" response header on a failed request: if the "server" response header includes "Airplay", then Airplay is the culprit.