Bridgy developer documentation

Bridgy connects your web site to social media. Likes, retweets, mentions, cross-posting, and more. See the user docs for more details, or the developer docs if you want to contribute.

https://brid.gy/

Bridgy is part of the IndieWeb ecosystem. In IndieWeb terminology, Bridgy offers backfeed, POSSE, and webmention support as a service.

License: This project is placed in the public domain.

Development

You’ll need the Google Cloud SDK (aka gcloud) with the gcloud-appengine-python and gcloud-appengine-python-extras components. Then, create a Python 3 virtualenv and install the dependencies with:

python3 -m venv local3
source local3/bin/activate
pip install -r requirements.txt

Now, run the unit tests:

gcloud beta emulators datastore start --no-store-on-disk --consistency=1.0 --host-port=localhost:8089 < /dev/null >& /dev/null
python3 -m unittest discover
kill %1

If you send a pull request, please include or update a test for your new code!

To run the entire app locally, run this in the repo root directory:

dev_appserver.py --log_level debug --enable_host_checking false \
  --support_datastore_emulator --datastore_emulator_port=8089 \
  --application=brid-gy ~/src/bridgy/app.yaml ~/src/bridgy/background.yaml

Open localhost:8080 and you should see the Bridgy home page!

If you hit an error during setup, check out the oauth-dropins Troubleshooting/FAQ section. For searchability, here are a handful of error messages that have solutions there:

bash: ./bin/easy_install: ...bad interpreter: No such file or directory

ImportError: cannot import name certs

ImportError: No module named dev_appserver

ImportError: cannot import name tweepy

File ".../site-packages/tweepy/auth.py", line 68, in _get_request_token
  raise TweepError(e)
TweepError: must be _socket.socket, not socket

error: option --home not recognized

There’s a good chance you’ll need to make changes to granary, oauth-dropins, or webmention-tools at the same time as bridgy. To do that, clone their repos elsewhere, then install them in “source” mode with:

pip uninstall -y oauth-dropins
pip install -e <path to oauth-dropins>

pip uninstall -y granary
pip install -e <path to granary>

pip uninstall -y webmentiontools
pip install <path to webmention-tools>

To deploy to App Engine, run scripts/deploy.sh.

remote_api_shell is a useful interactive Python shell that can interact with the production app’s datastore, memcache, etc. To use it, create a service account and download its JSON credentials, put it somewhere safe, and put its path in your GOOGLE_APPLICATION_CREDENTIALS environment variable.

Adding a new silo

So you want to add a new silo? Maybe MySpace, or Friendster, or even Tinder? Great! Here are the steps to do it. It looks like a lot, but it’s not that bad, honest.

  1. Find the silo’s API docs and check that it can do what Bridgy needs. At minimum, it should be able to get a user’s posts and their comments, likes, and reposts, depending on which of those the silo supports. If you want publish support, it should also be able to create posts, comments, likes, reposts, and/or RSVPs.

  2. Fork and clone this repo.

  3. Create an app (aka client) in the silo’s developer console, grab your app’s id (aka key) and secret, put them into new local files in the repo root dir, following this pattern. You’ll eventually want to send them to @snarfed too, but no hurry.

  4. Add the silo to oauth-dropins if it’s not already there:

    1. Add a new .py file for your silo with an auth model and handler classes. Follow the existing examples.

    2. Add a 100 pixel tall button image named [NAME]_2x.png, where [NAME] is your start handler class’s NAME constant, eg 'twitter'.

    3. Add it to the app front page and the README.

  5. Add the silo to granary:

    1. Add a new .py file for your silo. Follow the existing examples. At minimum, you’ll need to implement get_activities_response and convert your silo’s API data to ActivityStreams.

    2. Add a new unit test file and write some tests!

    3. Add it to api.py (specifically Handler.get), app.py, index.html, and the README.

  6. Add the silo to Bridgy:

    1. Add a new .py file for your silo with a model class. Follow the existing examples.

    2. Add it to app.py and handlers.py (just import the module).

    3. Add a 48x48 PNG icon to static/.

    4. Add a new [SILO]_user.html file in templates/ and add the silo to index.html. Follow the existing examples.

    5. Add the silo to about.html and this README.

    6. If users’ profile picture URLs can change, add a cron job that updates them to cron.py.

  7. Optionally add publish support:

    1. Implement create and preview_create for the silo in granary.

    2. Add the silo to publish.py: import its module, add it to SOURCES, and update this error message.

Good luck, and happy hacking!

Monitoring

App Engine’s built in dashboard and log browser are pretty good for interactive monitoring and debugging.

For alerting, we’ve set up Google Cloud Monitoring (née Stackdriver). Background in issue 377. It sends alerts by email and SMS when HTTP 4xx responses average >.1qps or 5xx >.05qps, latency averages >15s, or instance count averages >5 over the last 15m window.

Stats

I occasionally generate stats and graphs of usage and growth from the BigQuery dataset (#715). Here’s how.

  1. Export the full datastore to Google Cloud Storage. Include all entities except *Auth and other internal details. Check to see if any new kinds have been added since the last time this command was run.

    gcloud datastore export --async gs://brid-gy.appspot.com/stats/ --kinds Blogger,BlogPost,BlogWebmention,FacebookPage,Flickr,GitHub,GooglePlusPage,Instagram,Medium,Publish,PublishedPage,Response,SyndicatedPost,Tumblr,Twitter,WordPress
    

    Note that --kinds is required. From the export docs, Data exported without specifying an entity filter cannot be loaded into BigQuery.

  2. Wait for it to be done with gcloud datastore operations list | grep done.

  3. Import it into BigQuery:

    for kind in BlogPost BlogWebmention Publish Response SyndicatedPost; do
      bq load --replace --nosync --source_format=DATASTORE_BACKUP datastore.$kind gs://brid-gy.appspot.com/stats/all_namespaces/kind_$kind/all_namespaces_kind_$kind.export_metadata
    done
    
    for kind in Blogger FacebookPage Flickr GitHub GooglePlusPage Instagram Medium Meetup Tumblr Twitter WordPress; do
      bq load --replace --nosync --source_format=DATASTORE_BACKUP sources.$kind gs://brid-gy.appspot.com/stats/all_namespaces/kind_$kind/all_namespaces_kind_$kind.export_metadata
    done
    
  4. Check the jobs with bq ls -j, then wait for them with bq wait.

  5. Run the full stats BigQuery query. Download the results as CSV.

  6. Open the stats spreadsheet. Import the CSV, replacing the data sheet.

  7. Check out the graphs! Save full size images with OS or browser screenshots, thumbnails with the Save Image button. Then post them!

Misc

The datastore is automatically backed up by an App Engine cron job that runs Datastore managed export (details) and stores the results in Cloud Storage, in the brid-gy.appspot.com bucket. It backs up weekly and includes all entities except Response and SyndicatedPost, since they make up 92% of all entities by size and they aren’t as critical to keep.

(We used to use Datastore Admin Backup, but it shut down in Feb 2019.)

We use this command to set a Cloud Storage lifecycle policy on that bucket that prunes older backups:

gsutil lifecycle set cloud_storage_lifecycle.json gs://brid-gy.appspot.com

Run this to see how much space we’re currently using:

gsutil du -hsc gs://brid-gy.appspot.com/\*

Run this to download a single complete backup:

gsutil -m cp -r gs://brid-gy.appspot.com/weekly/datastore_backup_full_YYYY_MM_DD_\* .

Also see the BigQuery dataset (#715).