This is an implementation of a Protobuf schema registry. Right now, the implementation is fairly basic and focuses on the documentation aspects of a registry. It is used internally at Spotify to power our API documentation for gRPC services.
A test instance is hosted on Netlify at https://spotify-proto-registry.netlify.com/. Note that this instance only hosts documentation, and will not serve Type requests correctly, because Netlify only offers hosting of static HTML pages.
This repository builds a docker image that can be used to generate a Protobuf registry for any set of schemas.
For example, using Google Cloud Build, put this in your cloudbuild.yaml
:
steps:
- name: 'spotify/proto-registry-builder:<version>'
envs:
INPUT: 'schema.fds.pb' # path to a binary Protobuf file containing a FileDescriptorSet
OUTPUT: 'registry' # path to a directory where a Dockerfile + data should be put
- name: 'gcr.io/cloud-builders/docker'
args: [ 'build', '-t', 'gcr.io/$PROJECT_ID/registry', 'registry' ]
Right now, the building process is a bit hacky. It assumes that there is a file in
src/schema/schema.pb
containing a FileDescriptorSet
containing the schema to be browsed. You
can generate that file by running this command on some arbitrary protobufs:
protoc --include_imports --include_source_info \
-o path/to/src/schema/schema.pb \
-I dir dir/**/*.proto
There is a file descriptor set used by the tests that is generated with the ./test/update-testdata
script that you can use if you don't have protoc
installed:
gunzip -k test/testdata.fds.pb.gz
cp test/testdata.fds.pb src/schema/schema.pb
After that, simply run the usual:
yarn
yarn start
The registry responds to resources at /<type name>
only. This matches the API of the
Any
type. The idea is that the registry should respond with documentation when
requested (e.g. when the request specifies Accept: text/html
or similar) and with a
Type
otherwise.
This project adheres to the Open Code of Conduct. By participating, you are expected to honor this code.