Changelog

You can subscribe to release announcements by:

This changelog is viewable on the web at https://docs.funkwhale.audio/changelog.html.

0.16.3 (2018-08-21)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Bugfixes:

  • Fixed front-end not contacting the proper path on the API (!385)

0.16.2 (2018-08-21)

Warning

This release is broken, do not use it. Upgrade to 0.16.3 or higher instead.

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Bugfixes:

  • Ensure we always have a default api url set on first load to avoid displaying the instance picker (#490)
  • Fixed CLI importer syntax error because of async reserved keyword usage (#494)

0.16.1 (2018-08-19)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Features:

  • Make funkwhale themable by loading external stylesheets (#456)

Enhancements:

  • Add link to admin on “Staff member” button (#202)
  • Can now add a description to radios and better radio cards (#331)
  • Display track duration in track tables (#461)
  • More permissive default permissions for front-end files (#388)
  • Simpler configuration and toolchain for the front-end using vue-cli (!375)
  • Use Howler to manage audio instead of our own dirty/untested code (#392)

Bugfixes:

  • Fix alignment issue on top bar in Admin tabs (#395)
  • Fix Apache2 permission issue preventing /media folder from being served correctly (#389)
  • Fix loading on browse page lists causing them to go down, and dimming over the top bar (#468)
  • Fixed (again): administration section not showing up in sidebar after login (#245)
  • Fixed audio mimetype not showing up on track detail and list (#459)
  • Fixed broken audio playback on Chrome and invisible volume control (#390)
  • Fixed broken federation import on big imports due to missing transaction logic (#397)
  • Fixed crash on artist pages when no cover is available (#457)
  • Fixed favorited status of tracks not appearing in interface (#398)
  • Fixed invitation code not prefilled in form when accessing invitation link (#476)
  • Fixed typos in scheduled tasks configuration (#487)
  • Removed release date error in case of empty date (#478)
  • Removed white on white artist button on hover, on Album page (#393)
  • Smarter date parsing during import by replacing arrow with pendulum (#376)
  • Display public playlists properly for anonymous users (#488)

i18n:

  • Added portuguese, spanish and german translations

Custom themes for Funkwhale

If you ever wanted to give a custom look and feel to your instance, this is now possible.

Check https://docs.funkwhale.audio/configuration.html#theming if you want to know more!

Fix Apache2 configuration file for media block [Manual action required]

The permission scope on the current Apache2 configuration file is too narrow, preventing thumbnails from being served.

On Apache2 setups, you have to replace the following line:

<Directory /srv/funkwhale/data/media/albums>

with:

<Directory /srv/funkwhale/data/media>

You can now restart your server:

sudo systemctl restart apache2

0.16 (2018-07-22)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Features:

  • Complete redesign of the library home and playlist pages (#284)
  • Expose ActivityPub actors for users (#317)
  • Implemented a basic but functionnal Github-like search on federated tracks list (#344)
  • Internationalized interface as well as translations for Arabic, French, Esperanto, Italian, Occitan, Polish, Portuguese and Swedish (#161, #167)
  • Users can now upload an avatar in their settings page (#257)

Enhancements:

  • Added feedback when creating/updating radio (#302)
  • Apply restrictions to username characters during signup
  • Autoselect best language based on browser configuration (#386)
  • Can now order tracks on federated track list (#326)
  • Can now relaunch pending import jobs from the web interface (#323)
  • Ensure we do not display pagination on single pages (#334)
  • Ensure we have sane defaults for MEDIA_ROOT, STATIC_ROOT and MUSIC_DIRECTORY_PATH in the deployment .env file (#350)
  • Make some space for the volume slider to allow precise control (#318)
  • Removed django-cacheops dependency
  • Store track artist and album artist separately (#237) Better handling of tracks with a different artist than the album artist
  • The navigation bar of Library is now fixed (#375)
  • Use thumbnails for avatars and covers to reduce bandwidth

Bugfixes:

  • Ensure 750 permissions on CI artifacts (#332)
  • Ensure images are not cropped in queue (#337)
  • Ensure we do not import artists with empty names (#351)
  • Fix notifications not closing when clicking on the cross (#366)
  • Fix the most annoying offset in the whole fediverse (#369)
  • Fixed persistent message in playlist modal (#304)
  • Fixed unfiltered results in favorites API (#384)
  • Raise a warning instead of crashing when getting a broken path in file import (#138)
  • Remove parallelization of uploads during import to avoid crashing small servers (#382)
  • Subsonic API login is now case insensitive (#339)
  • Validate Date header in HTTP Signatures (#328)

Documentation:

  • Added troubleshotting and technical overview documentation (#256)
  • Arch Linux installation steps
  • Document that users can use Ultrasonic on Android (#316)
  • Fixed a couple of typos
  • Some cosmetic improvements to the doc

i18n:

  • Arabic translation (!302)
  • Polish translation (!304)

Library home and playlist page overhaul

The library home page have been completely redesigned to include:

  • other users activity (listenings, playlists and favorites)
  • recently imported albums

We think this new version showcases more music in a more useful way, let us know what you think about it!

The playlist page have been updated as well.

Internationalized interface

After months of work, we’re proud to announce our interface is now ready for internationalization.

Translators have already started the work of translating Funkwhale in 8 different languages, and we’re ready to add more as needed.

You can easily get involved at https://translate.funkwhale.audio/engage/funkwhale/

Better handling of tracks with a different artist than the album artist

Some tracks involve a different artist than the album artist (e.g. a featuring) and Funkwhale has been known to do weird things when importing such tracks, resulting in albums that contained a single track, for instance.

The situation should be improved with this release, as Funkwhale is now able to store separately the track and album artist, and display it properly in the interface.

Users now have an ActivityPub Actor [Manual action required]

In the process of implementing federation for user activity such as listening history, we are now making user profiles (a.k.a. ActivityPub actors) available through federation.

This does not means the federation is working, but this is a needed step to implement it.

Those profiles will be created automatically for new users, but you have to run a command to create them for existing users.

On docker setups:

docker-compose run --rm api python manage.py script create_actors --no-input

On non-docker setups:

python manage.py script create_actors --no-input

This should only take a few seconds to run. It is safe to interrupt the process or rerun it multiple times.

Image thumbnails [Manual action required]

To reduce bandwidth usage on slow or limited connexions and improve performance in general, we now use smaller images in the front-end. For instance, if you have an album cover with a 1000x1000 pixel size, we will create smaller versions of this image (50x50, 200x200, 400x400) and reference those resized version when we don’t actually need the original image.

Thumbnail will be created automatically for new objects, however, you have to launch a manual command to deal with existing ones.

On docker setups:

docker-compose run --rm api python manage.py script create_image_variations --no-input

On non-docker setups:

python manage.py script create_image_variations --no-input

This should be quite fast but may take up to a few minutes depending on the number of albums you have in database. It is safe to interrupt the process or rerun it multiple times.

Improved search on federated tracks list

Having a powerful but easy-to-use search is important but difficult to achieve, especially if you do not want to have a real complex search interface.

Github does a pretty good job with that, using a structured but simple query system (See https://help.github.com/articles/searching-issues-and-pull-requests/#search-only-issues-or-pull-requests).

This release implements a limited but working subset of this query system. You can use it only on the federated tracks list (/manage/federation/tracks) at the moment, but depending on feedback it will be rolled-out on other pages as well.

This is the type of query you can run:

  • hello world: search for “hello” and “world” in all the available fields
  • hello in:artist search for results where artist name is “hello”
  • spring in:artist,album search for results where artist name or album title contain “spring”
  • artist:hello search for results where artist name equals “hello”
  • artist:"System of a Down" domain:instance.funkwhale search for results where artist name equals “System of a Down” and inside “instance.funkwhale” library

Ensure MEDIA_ROOT, STATIC_ROOT and MUSIC_DIRECTORY_* are set explicitely [Manual action required]

In our default .env file, MEDIA_ROOT and STATIC_ROOT were commented by default, causing some deployment issues on non-docker setups when people forgot to uncomment them.

From now on, those variables are uncommented, and will also be used on docker setups to mount the volumes automatically in the docker-compose.yml file. This has been a source of headache as well in some deployments, where you had to update both the .env file and the compose file.

This also applies to in-place paths (MUSIC_DIRECTORY_PATH and MUSIC_DIRECTORY_SERVE_PATH), whose values are now used directly to set up the proper Docker volumes.

This will only affect new deployments though. If you want to benefit from this on an existing instance, do a backup of your .env and docker-compose.yml files and apply the following changes:

  • Ensure MEDIA_ROOT is uncommented in your .env file and match the absolute path where media files are stored on your host (/srv/funkwhale/data/media by default)
  • Ensure STATIC_ROOT is uncommented in your .env file and match the absolute path where static files are stored on your host (/srv/funkwhale/data/static by default)
  • If you use in-place import:
    • Ensure MUSIC_DIRECTORY_PATH is uncommented and set to /music
    • Ensure MUSIC_DIRECTORY_SERVE_PATH is uncommented and set to the absolute path on your host were your music files are stored (/srv/funkwhale/data/music by default)
  • Edit your docker-compose.yml file to reflect the changes:
    • Search for volumes (there should be two occurences) that contains /app/funkwhale_api/media on the right side, and replace the whole line with - "${MEDIA_ROOT}:${MEDIA_ROOT}"
    • Search for a volume that contains /app/staticfiles on the right side, and replace the whole line with - "${STATIC_ROOT}:${STATIC_ROOT}"
    • If you use in-place import, search for volumes (there should be two occurences) that contains /music:ro on the right side, and replace the whole line with - "${MUSIC_DIRECTORY_SERVE_PATH}:${MUSIC_DIRECTORY_PATH}:ro"

In the end, the volumes directives of your containers should look like that:

...
celeryworker
  volumes:
    - "${MUSIC_DIRECTORY_SERVE_PATH}:${MUSIC_DIRECTORY_PATH}:ro"
    - "${MEDIA_ROOT}:${MEDIA_ROOT}"
...
api:
  volumes:
    - "${MUSIC_DIRECTORY_SERVE_PATH}:${MUSIC_DIRECTORY_PATH}:ro"
    - "${MEDIA_ROOT}:${MEDIA_ROOT}"
    - "${STATIC_ROOT}:${STATIC_ROOT}"
    - ./front/dist:/frontend
...

Removed Cacheops dependency

We removed one of our dependency named django-cacheops. It was unly used in a few places, and not playing nice with other dependencies.

You can safely remove this dependency in your environment with pip uninstall django-cacheops if you’re not using docker.

You can also safely remove any CACHEOPS_ENABLED setting from your environment file.

0.15 (2018-06-24)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Features:

  • Added admin interface to manage import requests (#190)
  • Added replace flag during import to replace already present tracks with a new version of their track file (#222)
  • Funkwhale’s front-end can now point to any instance (#327) Removed front-end and back-end coupling
  • Management interface for users (#212)
  • New invite system (#248) New invite system

Enhancements:

  • Added “TV” to the list of highlighted words during YouTube import (#154)
  • Command line import now accepts unlimited args (#242)

Bugfixes:

  • Expose track files date in manage API (#307)
  • Fixed current track restart/hiccup when shuffling queue, deleting track from queue or reordering (#310)
  • Include user’s current private playlists on playlist list (#302)
  • Remove link to generic radios, since they don’t have detail pages (#324)

Documentation:

  • Document that Funkwhale may be installed with YunoHost (#325)
  • Documented a saner layout with symlinks for in-place imports (#254)
  • Upgrade documentation now use the correct user on non-docker setups (#265)

Invite system

On closed instances, it has always been a little bit painful to create accounts by hand for new users. This release solve that by adding invitations.

You can generate invitation codes via the “users” admin interface (you’ll find a link in the sidebar). Those codes are valid for 14 days, and can be used once to create a new account on the instance, even if registrations are closed.

By default, we generate a random code for invitations, but you can also use custom codes if you need to print them or make them fancier ;)

Invitations generation and management requires the “settings” permission.

Removed front-end and back-end coupling

Eventhough Funkwhale’s front-end has always been a Single Page Application, talking to an API, it was only able to talk to an API on the same domain.

There was no real technical justification behind this (only lazyness), and it was also blocking interesting use cases:

  • Use multiple customized versions of the front-end with the same instance
  • Use a customized version of the front-end with multiple instances
  • Use a locally hosted front-end with a remote API, which is especially useful in development

From now on, Funkwhale’s front-end can connect to any Funkwhale server. You can change the server you are connecting to in the footer.

Fixing this also unlocked a really interesting feature in our development/review workflow: by leveraging Gitlab CI and review apps, we are now able to deploy automatically live versions of a merge request, making it possible for anyone to review front-end changes easily, without the need to install a local environment.

0.14.2 (2018-06-16)

Warning

This release contains a fix for a permission issue. You should upgrade as soon as possible. Read the changelog below for more details.

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Enhancements:

  • Added feedback on shuffle button (#262)
  • Added multiple warnings in the documentation that you should never run makemigrations yourself (#291)
  • Album cover served in http (#264)
  • Apache2 reverse proxy now supports websockets (tested with Apache 2.4.25) (!252)
  • Display file size in human format during file upload (#289)
  • Switch from BSD-3 licence to AGPL-3 licence (#280)

Bugfixes:

  • Ensure radios can only be edited and deleted by their owners (#311)
  • Fixed admin menu not showing after login (#245)
  • Fixed broken pagination in Subsonic API (#295)
  • Fixed duplicated websocket connexion on timeline (#287)

Documentation:

  • Improved documentation about in-place imports setup (#298)

Other:

  • Added Black and flake8 checks in CI to ensure consistent code styling and formatting (#297)
  • Added bug and feature issue templates (#299)

Permission issues on radios

Because of an error in the way we checked user permissions on radios, public radios could be deleted by any logged-in user, even if they were not the owner of the radio.

We recommend instances owners to upgrade as fast as possible to avoid any abuse and data loss.

Funkwhale is now licenced under AGPL-3

Following the recent switch made by PixelFed (https://github.com/dansup/pixelfed/issues/143), we decided along with the community to relicence Funkwhale under the AGPL-3 licence. We did this switch for various reasons:

  • This is better aligned with other fediverse software
  • It prohibits anyone to distribute closed-source and proprietary forks of Funkwhale

As end users and instance owners, this does not change anything. You can continue to use Funkwhale exactly as you did before :)

Apache support for websocket

Up until now, our Apache2 configuration was not working with websockets. This is now solved by adding this at the beginning of your Apache2 configuration file:

Define funkwhale-api-ws ws://localhost:5000

And this, before the “/api” block:

# Activating WebSockets
ProxyPass "/api/v1/instance/activity" ${funkwhale-api-ws}/api/v1/instance/activity

Websockets may not be supported in older versions of Apache2. Be sure to upgrade to the latest version available.

Serving album covers in https (Apache2 proxy)

Two issues are addressed here. The first one was about Django replying with mixed content (http) when queried for covers. Setting up the X-Forwarded-Proto allows Django to know that the client is using https, and that the reply must be https as well.

Second issue was a problem of permission causing Apache a denied access to album cover folder. It is solved by adding another block for this path in the Apache configuration file for funkwhale.

Here is how to modify your funkwhale.conf apache2 configuration:

<VirtualHost *:443>

  ...
  #Add this new line
  RequestHeader set X-Forwarded-Proto "https"
  ...
  # Add this new block below the other <Directory/> blocks
  # replace /srv/funkwhale/data/media with the path to your media directory
  # if you're not using the standard layout.
  <Directory /srv/funkwhale/data/media/albums>
    Options FollowSymLinks
    AllowOverride None
    Require all granted
  </Directory>
  ...
</VirtualHost>

About the makemigrations warning

You may sometimes get the following warning while applying migrations:

"Your models have changes that are not yet reflected in a migration, and so won't be applied."

This is a warning, not an error, and it can be safely ignored. Never run the makemigrations command yourself.

0.14.1 (2018-06-06)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Enhancements:

  • Display server version in the footer (#270)
  • fix_track_files will now update files with bad mimetype (and not only the one with no mimetype) (#273)
  • Huge performance boost (~x5 to x7) during CLI import that queries MusicBrainz (#288)
  • Removed alpha-state transcoding support (#271)

Bugfixes:

  • Broken logging statement during import error (#274)
  • Broken search bar on library home (#278)
  • Do not crash when importing track with an artist that do not match the release artist (#237)
  • Do not crash when tag contains multiple uuids with a / separator (#267)
  • Ensure we do not store bad mimetypes (such as application/x-empty) (#266)
  • Fix broken “play all” button that played only 25 tracks (#281)
  • Fixed broken track download modal (overflow and wrong URL) (#239)
  • Removed hardcoded size limit in file upload widget (#275)

Documentation:

  • Added warning about _protected/music location in nginx configuration (#247)

Removed alpha-state transcoding (#271)

A few months ago, a basic transcoding feature was implemented. Due to the way this feature was designed, it was slow, CPU intensive on the server side, and very tightly coupled to the reverse-proxy configuration, preventing it to work Apache2, for instance. It was also not compatible with Subsonic clients.

Based on that, we’re currently removing support for transcoding in its current state. The work on a better designed transcoding feature can be tracked in https://code.eliotberriot.com/funkwhale/funkwhale/issues/272.

You don’t have to do anything on your side, but you may want to remove the now obsolete configuration from your reverse proxy file (nginx only):

# Remove those blocks:

# transcode cache
proxy_cache_path /tmp/funkwhale-transcode levels=1:2 keys_zone=transcode:10m max_size=1g inactive=7d;

# Transcoding logic and caching
location = /transcode-auth {
    include /etc/nginx/funkwhale_proxy.conf;
    # needed so we can authenticate transcode requests, but still
    # cache the result
    internal;
    set $query '';
    # ensure we actually pass the jwt to the underlytin auth url
    if ($request_uri ~* "[^\?]+\?(.*)$") {
        set $query $1;
    }
    proxy_pass http://funkwhale-api/api/v1/trackfiles/viewable/?$query;
    proxy_pass_request_body off;
    proxy_set_header        Content-Length "";
}

location /api/v1/trackfiles/transcode/ {
    include /etc/nginx/funkwhale_proxy.conf;
    # this block deals with authenticating and caching transcoding
    # requests. Caching is heavily recommended as transcoding
    # is a CPU intensive process.
    auth_request /transcode-auth;
    if ($args ~ (.*)jwt=[^&]*(.*)) {
        set $cleaned_args $1$2;
    }
    proxy_cache_key "$scheme$request_method$host$uri$is_args$cleaned_args";
    proxy_cache transcode;
    proxy_cache_valid 200 7d;
    proxy_ignore_headers "Set-Cookie";
    proxy_hide_header "Set-Cookie";
    add_header X-Cache-Status $upstream_cache_status;
    proxy_pass   http://funkwhale-api;
}
# end of transcoding logic

0.14 (2018-06-02)

Upgrade instructions are available at
https://docs.funkwhale.audio/upgrading.html

Features:

  • Admins can now configure default permissions that will be granted to all registered users (#236)
  • Files management interface for users with “library” permission (#223)
  • New action table component for quick and efficient batch actions (#228) This is implemented on the federated tracks pages, but will be included in other pages as well depending on the feedback.

Enhancements:

  • Added a new “upload” permission that allows user to launch import and view their own imports (#230)
  • Added Support for OggTheora in import.
  • Autoremove media files on model instance deletion (#241)
  • Can now import a whole remote library at once thanks to new Action Table component (#164)
  • Can now use album covers from flac/mp3 metadata and separate file in track directory (#219)
  • Implemented getCovertArt in Subsonic API to serve album covers (#258)
  • Implemented scrobble endpoint of subsonic API, listenings are now tracked correctly from third party apps that use this endpoint (#260)
  • Retructured music API to increase performance and remove useless endpoints (#224)

Bugfixes:

  • Consistent constraints/checks for URL size (#207)
  • Display proper total number of tracks on radio detail (#225)
  • Do not crash on flac import if musicbrainz tags are missing (#214)
  • Empty save button in radio builder (#226)
  • Ensure anonymous users can use the app if the instance is configured accordingly (#229)
  • Ensure inactive users cannot get auth tokens (#218) This was already the case bug we missed some checks
  • File-upload import now supports Flac files (#213)
  • File-upload importer should now work properly, assuming files are tagged (#106)
  • Fixed a few broken translations strings (#227)
  • Fixed broken ordering in front-end lists (#179)
  • Fixed ignored page_size paremeter on artist and favorites list (#240)
  • Read ID3Tag Tracknumber from TRCK (#220)
  • We now fetch album covers regardless of the import methods (#231)

Documentation:

  • Added missing subsonic configuration block in deployment vhost files (#249)
  • Moved upgrade doc under install doc in TOC (#251)

Other:

  • Removed acoustid support, as the integration was buggy and error-prone (#106)

Files management interface

This is the first bit of an ongoing work that will span several releases, to bring more powerful library management features to Funkwhale. This iteration includes a basic file management interface where users with the “library” permission can list and search available files, order them using various criterias (size, bitrate, duration…) and delete them.

New “upload” permission

This new permission is helpful if you want to give upload/import rights to some users, but don’t want them to be able to manage the library as a whole: although there are no controls yet for managing library in the fron-end, subsequent release will introduce management interfaces for artists, files, etc.

Because of that, users with the “library” permission will have much more power, and will also be able to remove content from the platform. On the other hand, users with the “upload” permission will only have the ability to add new content.

Also, this release also includes a new feature called “default permissions”: those are permissions that are granted to every users on the platform. On public/open instances, this will play well with the “upload” permission since everyone will be able to contribute to the instance library without an admin giving the permission to every single user.

Smarter album cover importer

In earlier versions, covers where only imported when launching a YouTube import. Starting from this release, covers will be imported regardless of the import mode (file upload, youtube-dl, CLI, in-place…). Funkwhale will look for covers in the following order:

  1. In the imported file itself (FLAC/MP3 only)
  2. In a cover.jpg or cover.png in the file directory
  3. By fetching cover art from Musibrainz, assuming the file is tagged correctly

This will only work for newly imported tracks and albums though. In the future, we may offer an option to refetch album covers from the interface, but in the meantime, you can use the following snippet:

# Store this in /tmp/update_albums.py
from funkwhale_api.music.models import Album, TrackFile
from funkwhale_api.music.tasks import update_album_cover

albums_without_covers = Album.objects.filter(cover='')
total = albums_without_covers.count()
print('Found {} albums without cover'.format(total))
for i, album in enumerate(albums_without_covers.iterator()):
    print('[{}/{}] Fetching cover for {}...'.format(i+1, total, album.title))
    f = TrackFile.objects.filter(track__album=album).filter(source__startswith='file://').first()
    update_album_cover(album, track_file=f)

Then launch it:

# docker setups
cat /tmp/update_albums.py | docker-compose run --rm api python manage.py shell -i python

# non-docker setups
source /srv/funkwhale/load_env
source /srv/funkwhale/virtualenv/bin/activate
cat /tmp/update_albums.py | python manage.py shell -i python

# cleanup
rm /tmp/update_albums.py

Note

Depending on your number of albums, the previous snippet may take some time to execute. You can interrupt it at any time using ctrl-c and relaunch it later, as it’s idempotent.

Music API changes

This release includes an API break. Even though the API is advertised as unstable, and not documented, here is a brief explanation of the change in case you are using the API in a client or in a script. Summary of the changes:

  • /api/v1/artists does not includes a list of tracks anymore. It was to heavy to return all of this data all the time. You can get all tracks for an artist using /api/v1/tracks?artist=artist_id
  • Additionally, /api/v1/tracks now support an album filter to filter tracks matching an album
  • /api/v1/artists/search, /api/v1/albums/search and /api/v1/tracks/search endpoints are removed. Use /api/v1/{artists|albums|tracks}/?q=yourquery instead. It’s also more powerful, since you can combine search with other filters and ordering options.
  • /api/v1/requests/import-requests/search endpoint is removed as well. Use /api/v1/requests/import-requests/?q=yourquery instead. It’s also more powerful, since you can combine search with other filters and ordering options.

Of course, the front-end was updated to work with the new API, so this should not impact end-users in any way, apart from slight performance gains.

Note

The API is still not stable and may evolve again in the future. API freeze will come at a later point.

Flac files imports via upload

You have nothing to do to benefit from this, however, since Flac files tend to be a lot bigger than other files, you may want to increase the client_max_body_size value in your Nginx configuration if you plan to upload flac files.

Missing subsonic configuration bloc in vhost files

Because of a missing bloc in the sample Nginx and Apache configurations, instances that were deployed after the 0.13 release are likely to be unable to answer to Subsonic clients (the missing bits were properly documented in the changelog).

Ensure you have the following snippets in your Nginx or Apache configuration if you plan to use the Subsonic API.

Nginx:

location /rest/ {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/api/subsonic/rest/;
}

Apache2:

<Location "/rest">
    ProxyPass ${funkwhale-api}/api/subsonic/rest
    ProxyPassReverse ${funkwhale-api}/api/subsonic/rest
 </Location>

0.13 (2018-05-19)

Upgrade instructions are available at
https://docs.funkwhale.audio/upgrading.html

Features:

  • Can now import and play flac files (#157)
  • Simpler permission system (#152)
  • Store file length, size and bitrate (#195)
  • We now have a brand new instance settings interface in the front-end (#206)

Enhancements:

  • Disabled browsable HTML API in production (#205)
  • Instances can now indicate on the nodeinfo endpoint if they want to remain private (#200)

Bugfixes:

  • .well-known/nodeinfo endpoint can now answer to request with Accept: application/json (#197)
  • Fixed escaping issue of track name in playlist modal (#201)
  • Fixed missing dot when downloading file (#204)
  • In-place imported tracks with non-ascii characters don’t break reverse-proxy serving (#196)
  • Removed Python 3.6 dependency (secrets module) (#198)
  • Uplayable tracks are now properly disabled in the interface (#199)

Instance settings interface

Prior to this release, the only way to update instance settings (such as instance description, signup policy, federation configuration, etc.) was using the admin interface provided by Django (the back-end framework which power the API).

This interface worked, but was not really-user friendly and intuitive.

Starting from this release, we now offer a dedicated interface directly in the front-end. You can view and edit all your instance settings from here, assuming you have the required permissions.

This interface is available at ``/manage/settings` and via link in the sidebar.

Storage of bitrate, size and length in database

Starting with this release, when importing files, Funkwhale will store additional information about audio files:

  • Bitrate
  • Size (in bytes)
  • Duration

This change is not retroactive, meaning already imported files will lack those informations. The interface and API should work as before in such case, however, we offer a command to deal with legacy files and populate the missing values.

On docker setups:

docker-compose run --rm api python manage.py fix_track_files

On non-docker setups:

# from your activated virtualenv
python manage.py fix_track_files

Note

The execution time for this command is proportional to the number of audio files stored on your instance. This is because we need to read the files from disk to fetch the data. You can run it in the background while Funkwhale is up.

It’s also safe to interrupt this command and rerun it at a later point, or run it multiple times.

Use the –dry-run flag to check how many files would be impacted.

Simpler permission system

Starting from this release, the permission system is much simpler. Up until now, we were using Django’s built-in permission system, which was working, but also quite complex to deal with.

The new implementation relies on simpler logic, which will make integration on the front-end in upcoming releases faster and easier.

If you have manually given permissions to users on your instance, you can migrate those to the new system.

On docker setups:

docker-compose run --rm api python manage.py script django_permissions_to_user_permissions --no-input

On non-docker setups:

# in your virtualenv
python api/manage.py script django_permissions_to_user_permissions --no-input

There is still no dedicated interface to manage user permissions, but you can use the admin interface at /api/admin/users/user/ for that purpose in the meantime.

0.12 (2018-05-09)

Upgrade instructions are available at
https://docs.funkwhale.audio/upgrading.html

Features:

  • Subsonic API implementation to offer compatibility with existing clients such as DSub (#75)
  • Use nodeinfo standard for publishing instance information (#192)

Enhancements:

  • Play button now play tracks immediately instead of appending them to the queue (#99, #156)

Bugfixes:

  • Fix broken federated import (#193)

Documentation:

  • Up-to-date documentation for upgrading front-end files on docker setup (#132)

Subsonic API

This release implements some core parts of the Subsonic API, which is widely deployed in various projects and supported by numerous clients.

By offering this API in Funkwhale, we make it possible to access the instance library and listen to the music without from existing Subsonic clients, and without developping our own alternative clients for each and every platform.

Most advanced Subsonic clients support offline caching of music files, playlist management and search, which makes them well-suited for nomadic use.

Please head over Using Funkwhale from other apps for more informations about supported clients and user instructions.

At the instance-level, the Subsonic API is enabled by default, but require and additional endpoint to be added in you reverse-proxy configuration.

On nginx, add the following block:

location /rest/ {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/api/subsonic/rest/;
}

On Apache, add the following block:

<Location "/rest">
    ProxyPass ${funkwhale-api}/api/subsonic/rest
    ProxyPassReverse ${funkwhale-api}/api/subsonic/rest
</Location>

The Subsonic can be disabled at the instance level from the django admin.

Note

Because of Subsonic’s API design which assumes cleartext storing of user passwords, we chose to have a dedicated, separate password for that purpose. Users can generate this password from their settings page in the web client.

Nodeinfo standard for instance information and stats

Warning

The /api/v1/instance/stats/ endpoint which was used to display instance data in the about page is removed in favor of the new /api/v1/instance/nodeinfo/2.0/ endpoint.

In earlier version, we where using a custom endpoint and format for our instance information and statistics. While this was working, this was not compatible with anything else on the fediverse.

We now offer a nodeinfo 2.0 endpoint which provides, in a single place, all the instance information such as library and user activity statistics, public instance settings (description, registration and federation status, etc.).

We offer two settings to manage nodeinfo in your Funkwhale instance:

  1. One setting to completely disable nodeinfo, but this is not recommended as the exposed data may be needed to make some parts of the front-end work (especially the about page).
  2. One setting to disable only usage and library statistics in the nodeinfo endpoint. This is useful if you want the nodeinfo endpoint to work, but don’t feel comfortable sharing aggregated statistics about your library and user activity.

To make your instance fully compatible with the nodeinfo protocol, you need to to edit your nginx configuration file:

# before
...
location /.well-known/webfinger {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/.well-known/webfinger;
}
...

# after
...
location /.well-known/ {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/.well-known/;
}
...

You can do the same if you use apache:

# before
...
<Location "/.well-known/webfinger">
  ProxyPass ${funkwhale-api}/.well-known/webfinger
  ProxyPassReverse ${funkwhale-api}/.well-known/webfinger
</Location>
...

# after
...
<Location "/.well-known/">
  ProxyPass ${funkwhale-api}/.well-known/
  ProxyPassReverse ${funkwhale-api}/.well-known/
</Location>
...

This will ensure all well-known endpoints are proxied to funkwhale, and not just webfinger one.

Links:

0.11 (2018-05-06)

Upgrade instructions are available at https://docs.funkwhale.audio/upgrading.html

Special thanks for this release go to @renon:matrix.org (@Hazmo on Gitlab) for bringing Apache2 support to Funkwhale and contributing on other issues. Thank you!

Features:

Enhancements:

  • Added a fix_track_files command to run checks and fixes against library (#183)
  • Avoid fetching Actor object on every request authentication
  • Can now relaunch errored jobs and batches (#176)
  • List pending requests by default, added a status filter for requests (#109)
  • More structured menus in sidebar, added labels with notifications
  • Sample virtual-host file for Apache2 reverse-proxy (!165)
  • Store high-level settings (such as federation or auth-related ones) in database (#186)

Bugfixes:

  • Ensure in place imported files get a proper mimetype (#183)
  • Federation cache suppression is now simpler and also deletes orphaned files (#189)
  • Fixed small UI glitches/bugs in federation tabs (#184)
  • X-sendfile not working with in place import (#182)

Documentation:

  • Added a documentation area for third-party projects (#180)
  • Added documentation for optimizing Funkwhale and reduce its memory footprint.
  • Document that the database should use an utf-8 encoding (#185)
  • Foundations for API documentation with Swagger (#178)

Database storage for high-level settings

Due to the work done in #186, the following environment variables have been deprecated:

  • FEDERATION_ENABLED
  • FEDERATION_COLLECTION_PAGE_SIZE
  • FEDERATION_MUSIC_NEEDS_APPROVAL
  • FEDERATION_ACTOR_FETCH_DELAY
  • PLAYLISTS_MAX_TRACKS
  • API_AUTHENTICATION_REQUIRED

Configuration for this settings has been moved to database, as it will provide a better user-experience, by allowing you to edit these values on-the-fly, without restarting Funkwhale processes.

You can leave those environment variables in your .env file for now, as the values will be used to populate the database entries. We’ll make a proper announcement when the variables won’t be used anymore.

Please browse https://docs.funkwhale.audio/configuration.html#instance-settings for more information about instance configuration using the web interface.

System emails

Starting from this release, Funkwhale will send two types of emails:

  • Email confirmation emails, to ensure a user’s email is valid
  • Password reset emails, enabling user to reset their password without an admin’s intervention

Email sending is disabled by default, as it requires additional configuration. In this mode, emails are simply outputed on stdout.

If you want to actually send those emails to your users, you should edit your .env file and tweak the EMAIL_CONFIG variable. See EMAIL_CONFIG for more details.

Note

As a result of these changes, the DJANGO_EMAIL_BACKEND variable, which was not documented, has no effect anymore. You can safely remove it from your .env file if it is set.

Proxy headers for non-docker deployments

For non-docker deployments, add --proxy-headers at the end of the daphne command in /etc/systemd/system/funkwhale-server.service.

This will ensure the application receive the correct IP address from the client and not the proxy’s one.

0.10 (2018-04-23)

Features:

  • Can now import files in-place from the CLI importer (#155)

Enhancements:

  • Avoid downloading audio files multiple times from remote libraries (#163)
  • Better file import performance and error handling (#144)
  • Import job and batch API and front-end have been improved with better performance, pagination and additional filters (#171)
  • Increased max_length on TrackFile.source, this will help when importing files with a really long path (#142)
  • Player is back in Queue tab (#150)

Bugfixes:

  • Fail graciously when AP representation includes a null_value for mediaType
  • Fix sidebar tabs not showing under small resolution under Chrome (#173)
  • Fixed broken login due to badly configured Axios (#172)
  • Fixed broken playlist modal after login (#155)
  • Fixed queue reorder or track deletion restarting currently playing track (#151)
  • Radio will now append new track if you delete the last track in queue (#145)
  • Reset all sensitive front-end data on logout (#124)
  • Typos/not showing text due to i18n work (#175)

Documentation:

  • Better documentation for hardware requirements and memory usage (#165)

In-place import

This release includes in-place imports for the CLI import. This means you can load gigabytes of music into funkwhale without worrying about about Funkwhale copying those music files in its internal storage and eating your disk space.

This new feature is documented here and require additional configuration to ensure funkwhale and your webserver can serve those files properly.

Non-docker users:

Assuming your music is stored in /srv/funkwhale/data/music, add the following block to your nginx configuration:

location /_protected/music {
    internal;
    alias   /srv/funkwhale/data/music;
}

And the following to your .env file:

MUSIC_DIRECTORY_PATH=/srv/funkwhale/data/music

Docker users:

Assuming your music is stored in /srv/funkwhale/data/music, add the following block to your nginx configuration:

location /_protected/music {
    internal;
    alias   /srv/funkwhale/data/music;
}

Assuming you have the following volume directive in your docker-compose.yml (it’s the default): /srv/funkwhale/data/music:/music:ro, then add the following to your .env file:

# this is the path in the container
MUSIC_DIRECTORY_PATH=/music
# this is the path on the host
MUSIC_DIRECTORY_SERVE_PATH=/srv/funkwhale/data/music

0.9.1 (2018-04-17)

Bugfixes:

  • Allow null values for musicbrainz_id in Audio ActivityPub representation
  • Fixed broken permission check on library scanning and too aggressive page validation

0.9 (2018-04-17)

Features:

  • Add internationalization support (#5)
  • Can now follow and import music from remote libraries (#136, #137)

Enhancements:

  • Added a i18n-extract yarn script to extract strings to PO files (#162)
  • User admin now includes signup and last login dates (#148)
  • We now use a proper user agent including instance version and url during outgoing requests

Federation is here!

This is for real this time, and includes:

  • Following other Funkwhale libraries
  • Importing tracks from remote libraries (tracks are hotlinked, and only cached for a short amount of time)
  • Searching accross federated catalogs

Note that by default, federation is opt-in, on a per-instance basis: instances will request access to your catalog, and you can accept or refuse those requests. You can also revoke the access at any time.

Documentation was updated with relevant instructions to use and benefit from this new feature: https://docs.funkwhale.audio/federation.html

Preparing internationalization

Funkwhale’s front-end as always been english-only, and this is a barrier to new users. The work make Funkwhale’s interface translatable was started in this release by Baptiste. Although nothing is translated yet, this release includes behind the stage changes that will make it possible in the near future.

Many thank to Baptiste for the hard work and for figuring out a proper solution to this difficult problem.

Upgrade path

In addition to the usual instructions from https://docs.funkwhale.audio/upgrading.html, non-docker users will have to setup an additional systemd unit file for recurrent tasks.

This was forgotten in the deployment documentation, but recurrent tasks, managed by the celery beat process, will be needed more and more in subsequent releases. Right now, we’ll be using to clear the cache for federated music files and keep disk usage to a minimum.

In the future, they will also be needed to refetch music metadata or federated information periodically.

Celery beat can be enabled easily:

curl -L -o "/etc/systemd/system/funkwhale-beat.service" "https://code.eliotberriot.com/funkwhale/funkwhale/raw/develop/deploy/funkwhale-beat.service"
# Also edit /etc/systemd/system/funkwhale.target
# and ensure the Wants= line contains the following:
# Wants=funkwhale-server.service funkwhale-worker.service funkwhale-beat.service
nano /etc/systemd/system/funkwhale.target
# reload configuration
systemctl daemon-reload

Docker users already have celerybeat enabled.

0.8 (2018-04-02)

Features:

  • Add a detail page for radios (#64)
  • Implemented page title binding (#1)
  • Previous Track button restart playback after 3 seconds (#146)

Enhancements:

  • Added credits to Francis Gading for the logotype (#101)
  • API endpoint for fetching instance activity and updated timeline to use this new endpoint (#141)
  • Better error messages in case of missing environment variables (#140)
  • Implemented a @test@yourfunkwhaledomain bot to ensure federation works properly. Send it “/ping” and it will answer back :)
  • Queue shuffle now apply only to tracks after the current one (#97)
  • Removed player from queue tab and consistently show current track in queue (#131)
  • We now restrict some usernames from being used during signup (#139)

Bugfixes:

  • Better error handling during file import (#120)
  • Better handling of utf-8 filenames during file import (#138)
  • Converted favicon from .ico to .png (#130)
  • Upgraded to Python 3.6 to fix weird but harmless weakref error on django task (#121)

Documentation:

  • Documented the upgrade process (#127)

Preparing for federation

Federation of music libraries is one of the most asked feature. While there is still a lot of work to do, this version includes the foundation that will enable funkwhale servers to communicate between each others, and with other federated software, such as Mastodon.

Funkwhale will use ActivityPub as it’s federation protocol.

In order to prepare for federation (see #136 and #137), new API endpoints have been added under /federation and /.well-known/webfinger.

For these endpoints to work, you will need to update your nginx configuration, and add the following snippets:

location /federation/ {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/federation/;
}

location /.well-known/webfinger {
    include /etc/nginx/funkwhale_proxy.conf;
    proxy_pass   http://funkwhale-api/.well-known/webfinger;
}

This will ensure federation endpoints will be reachable in the future. You can of course skip this part if you know you will not federate your instance.

A new FEDERATION_ENABLED env var have also been added to control whether federation is enabled or not on the application side. This settings defaults to True, which should have no consequences at the moment, since actual federation is not implemented and the only available endpoints are for testing purposes.

Add FEDERATION_ENABLED=false to your .env file to disable federation on the application side.

To test and troubleshoot federation, we’ve added a bot account. This bot is available at @test@yourinstancedomain, and sending it “/ping”, for example, via Mastodon, should trigger a response.

0.7 (2018-03-21)

Features:

  • Can now filter artists and albums with no listenable tracks (#114)
  • Improve the style of the sidebar to make it easier to understand which tab is selected (#118)
  • On artist page, albums are not sorted by release date, if any (#116)
  • Playlists are here o/ :tada: (#3, #93, #94)
  • Use django-cacheops to cache common ORM requests (#117)

Bugfixes:

  • Fixed broken import request admin (#115)
  • Fixed forced redirection to login event with API_AUTHENTICATION_REQUIRED=False (#119)
  • Fixed position not being reseted properly when playing the same track multiple times in a row
  • Fixed synchronized start/stop radio buttons for all custom radios (#103)
  • Fixed typo and missing icon on homepage (#96)

Documentation:

  • Up-to-date and complete development and contribution instructions in README.rst (#123)

0.6.1 (2018-03-06)

Features:

  • Can now skip acoustid on file import with the –no-acoustid flag (#111)

Bugfixes:

  • Added missing batch id in output during import (#112)
  • Added some feedback on the play button (#100)
  • Smarter pagination which takes a fixed size (#84)

Other:

  • Completely removed django-cachalot from the codebase (#110). You can safely remove the CACHALOT_ENABLED setting from your .env file

0.6 (2018-03-04)

Features:

  • Basic activity stream for listening and favorites (#23)
  • Switched to django-channels and daphne for serving HTTP and websocket (#34)

This version contains breaking changes in the way funkwhale is deployed, please read the notes carefully.

Instance timeline

A new “Activity” page is now available from the sidebar, where you can browse your instance activity. At the moment, this includes other users favorites and listening, but more activity types will be implemented in the future.

Internally, we implemented those events by following the Activity Stream specification, which will help us to be compatible with other networks in the long-term.

A new settings page has been added to control the visibility of your activity. By default, your activity will be browsable by anyone on your instance, but you can switch to a full private mode where nothing is shared.

The setting form is available in your profile.

Switch from gunicorn to daphne

This release include an important change in the way we serve the HTTP API. To prepare for new realtime features and enable websocket support in Funkwhale, we are now using django-channels and daphne to serve HTTP and websocket traffic.

This replaces gunicorn and the switch should be easy assuming you follow the upgrade process described bellow.

If you are using docker, please remove the command instruction inside the api service, as the up-to-date command is now included directly in the image as the default entry point:

api:
  restart: unless-stopped
  image: funkwhale/funkwhale:${FUNKWHALE_VERSION:-latest}
  command: ./compose/django/gunicorn.sh  # You can remove this line

On non docker setups, you’ll have to update the [Service] block of your funkwhale-server systemd unit file to launch the application server using daphne instead of gunicorn.

The new configuration should be similar to this:

[Service]
User=funkwhale
# adapt this depending on the path of your funkwhale installation
WorkingDirectory=/srv/funkwhale/api
EnvironmentFile=/srv/funkwhale/config/.env
ExecStart=/usr/local/bin/daphne -b ${FUNKWHALE_API_IP} -p ${FUNKWHALE_API_PORT} config.asgi:application

Ensure you update funkwhale’s dependencies as usual to install the required packages.

On both docker and non-docker setup, you’ll also have to update your nginx configuration for websocket support. Ensure you have the following blocks included in your virtualhost file:

map $http_upgrade $connection_upgrade {
    default upgrade;
    ''      close;
}

server {
    ...
    proxy_http_version 1.1;
    proxy_set_header Upgrade $http_upgrade;
    proxy_set_header Connection $connection_upgrade;
}

Remember to reload your nginx server after the edit.

0.5.4 (2018-02-28)

Features:

  • Now stop running radio when clearing queue (#98)

Bugfixes:

  • Fixed queue skipping tracks (#91)
  • Now loop properly on queue when we only have one track (#95)

0.5.3 (2018-02-27)

Features:

  • Added admin interface for radios, track files, favorites and import requests (#80)
  • Added basic instance stats on /about (#82)
  • Search now unaccent letters for queries like “The Dø” or “Björk” yielding more results (#81)

Bugfixes:

  • Always use username in sidebar (#89)
  • Click event outside of player icons (#83)
  • Fixed broken import because of missing transaction
  • Now always load next radio track on last queue track ended (#87)
  • Now exclude tracks without file from radio candidates (#88)
  • skip to next track properly on 40X errors (#86)

Other:

  • Switched to towncrier for changelog management and compilation

0.5.2 (2018-02-26)

  • Fixed broken file import due to wrong url (#73)
  • More accurate mimetype detection
  • Fixed really small size on small screens
  • Added masonry layout for artists, requests and radios (#68)
  • We now have a favicon!
  • Fixed truncated play icon (#65)

0.5.1 (2018-02-24)

  • Front: Fixed broken ajax call on radio builder (#69)
  • Front: Shuffle now restart next track from beginning (#70)
  • Front: volume slider should now have the same style everywhere (#72)

0.5 (2018-02-24)

  • Front: Now reset player colors when track has no cover (#46)
  • Front: play button now disabled for unplayable tracks
  • API: You can now enable or disable registration on the fly, via a preference (#58)
  • Front: can now signup via the web interface (#35)
  • Front: Fixed broken redirection on login
  • Front: Fixed broken error handling on settings and login form

About page:

There is a brand new about page on instances (/about), and instance owner can now provide a name, a short and a long description for their instance via the admin (/api/admin/dynamic_preferences/globalpreferencemodel/).

Transcoding:

Basic transcoding is now available to/from the following formats : ogg and mp3.

This is still an alpha feature at the moment, please report any bug.

This relies internally on FFMPEG and can put some load on your server. It’s definitely recommended you setup some caching for the transcoded files at your webserver level. Check the the exemple nginx file at deploy/nginx.conf for an implementation.

On the frontend, usage of transcoding should be transparent in the player.

Music Requests:

This release includes a new feature, music requests, which allows users to request music they’d like to see imported. Admins can browse those requests and mark them as completed when an import is made.

0.4 (2018-02-18)

  • Front: ambiant colors in player based on current track cover (#59)
  • Front: simplified front dev setup thanks to webpack proxy (#59)
  • Front: added some unittests for the store (#55)
  • Front: fixed broken login redirection when 401
  • Front: Removed autoplay on page reload
  • API: Added a /instance/settings endpoint
  • Front: load /instance/settings on page load
  • Added settings to report JS and Python error to a Sentry instance This is disabled by default, but feel free to enable it if you want to help us by sending your error reports :) (#8)

0.3.5 (2018-01-07)

  • Smarter BACKEND_URL in frontend

0.3.4 (2018-01-07)

  • Fixed wrong URL construction in ajax call

0.3.3 (2018-01-07)

  • Users can now create their own dynamic radios (#51)

0.3.2

  • Fixed an issue in the main dockerfile

0.3.1

  • Revamped all import logic, everything is more tested and consistend
  • Can now use Acoustid in file imports to automatically grab metadata from musicbrainz
  • Brand new file import wizard

0.2.7

  • Shortcuts: can now use the f shortcut to toggle the currently playing track as a favorite (#53)
  • Shortcuts: avoid collisions between shortcuts by using the exact modifier (#53)
  • Player: Added looping controls and shortcuts (#52)
  • Player: Added shuffling controls and shortcuts (#52)
  • Favorites: can now modify the ordering of track list (#50)
  • Library: can now search/reorder results on artist browsing view (#50)
  • Upgraded celery to 4.1, added endpoint logic for fingerprinting audio files
  • Fixed #56: invalidate tokens on password change, also added change password form
  • Fixed #57: now refresh jwt token on page refresh
  • removed ugly dividers in batch import list
  • Fixed a few padding issues
  • Now persist/restore queue/radio/player state automatically
  • Removed old broken imports
  • Now force tests paths
  • Fixed #54: Now use pytest everywhere o/
  • Now use vuex to manage state for favorites
  • Now use vuex to manage state for authentication
  • Now use vuex to manage state for player/queue/radios

0.2.6 (2017-12-15)

  • Fixed broken Dockerfile

0.2.5 (2017-12-15)

Features:

  • Import: can now specify search template when querying import sources (#45)
  • Login form: now redirect to previous page after login (#2)
  • 404: a decent 404 template, at least (#48)

Bugfixes:

  • Player: better handling of errors when fetching the audio file (#46)
  • Csrf: default CSRF_TRUSTED_ORIGINS to ALLOWED_HOSTS to avoid Csrf issues on admin (#49)

Tech:

  • Django 2 compatibility, lot of packages upgrades (#47)

0.2.4 (2017-12-14)

Features:

  • Models: now store relese group mbid on Album model (#7)
  • Models: now bind import job to track files (#44)

Bugfixes:

  • Library: fixen broken “play all albums” button on artist cards in Artist browsing view (#43)