Bovine is a set of packages that simplify working with ActivityPub. The design goal is to provide an architecture that allows one to replace certain parts. For example the package bovine_store contains as the name indicates the data store being used. If the architecture was properly modular, it should be possible to replace bovine_store with a different implementation, without having to change anything in the rest of the code. I don't think I have met this goal yet, but in this sense bovine should be modular.
FIXME
Code is tested with pytest. Linted with flake8. Formatted with black. Furthermore some typing is enforced with mypy. Imports are sorted using isort. All the tests can be run via
./all_tests.shThe current codebase is split into packages with incrementally more complex use cases.
The bovine_core package contains the basic code necessary to build an ActivityPub client application.
A big caveat here is that currently authentication is assumed to be done through HTTP
signatures. It consists essentially of two parts:
bovine_core.clients.signed_http provides the functions signed_get and signed_post
which are used to communicate with the webserver, and bovine_core.activitystreams
contains builders used to construct the necessary ActivityStreams objects.
Two examples are included in the examples folder:
-
Munching cow demonstrates how to post simple content in
munching_cow.pyand then clean it back up inmunching_cow_cleanup.py. -
Cow Resisa provides a sample RSS to ActivityPub bridge.
The bovine package contains the server / client code to interact with ActivityPub. It uses asyncio using Quart and aiohttp.
A basic server implementation using it is available at examples/basic_app/basic_app.py. It should be noted that this example application does the absolute minimum. It only exposes what is needed to interact with ActivityPub. Also the user test is hard coded in it.
You can run it with
cd examples/basicapp
poetry run python basic_app.py
alternatively you can run it using hypercorn via
poetry run hypercorn basic_app:app
and then querying using
curl http://localhost:3335/.well-known/nodeinfo
curl http://localhost:3335/.well-known/webfinger?resource=acct:test
curl http://localhost:3335/activitypub/test
curl http://localhost:3335/activitypub/test -H accept:application/activity+json
By itself the bovine server does not do much: mostly just writeing things to stdout in reaction to requests. The bovine_tortoise package provides a database layer implemented with tortoise-orm. This allows for elements sent to the inbox to actually be recorded and for an outbox to be displayed. The database is currently managed with aerich.
The bovine_blog package provides an actual webserver on top of bovine + bovine_tortoise. Its look is still firmly basic. The HTML layouting and content presentation of the blog application can be developped independent of the rest of bovine by running:
cd examples/local_test_blog
poetry run hypercorn local_test_blog:app
The provided content is hardcoded in examples/local_test_blog/local_test_blog.py. Due to the nature of http signatures, the ids of posts correspond to the ActivityPub objects' canonical URI, e.g. https://my_domain/activitypub/test/1234-5678. The current implementation redirects requests to these urls not having the accept type matching application/.*json to https://my_domain/test/1234-5678.
bovine_blog/scripts contains some scripts that were useful during testing and development. In particular
poetry run python bovine_blog/scripts/add_user.py USERNAME
can be used to create new users including public/private keys.