[Feature Request] Subsonic API support

[balthild]

Eligibility

• I confirm that I am a genuine user and have purchased the software.
• Alternatively, I have paid for this feature request on Issue Hunt.

Feature Description

Describe the feature you'd like to request:

Subsonic API support.

• Streaming
• Playlist Managment

Use Case

Explain the use case for this feature:

• Navidrome
• Funkwhale
• Gonic
• ...

Expected Benefits

• Sync playlist across devices
• Serve music from a single source of library e.g., a server, an NAS, etc.

[Losses]

I'm not sure how many people will need this feature, so let's leave it here and see what the community thinks.

We are currently developing a complete and secure remote control protocol and media server for Rune, which I think should solve some requirements, I hope.

[balthild]

Would you accept if I want to contribute to it?

If yes, where could I start? I'm not very familiar to this codebase, but it seems that the library management is a separate crate decoupled from the rest of the app, and I could add the support to streaming services by adding an abstraction layer to that crate...

[Losses]

Would you accept if I want to contribute to it?

Thank you for your interest in contributing to Rune. Your proposal is valuable, and your contributing is welcomed.

While we welcome all contributions, I should note that Rune's query system has a fairly complex design that may require significant patience to understand fully. After reviewing the Subsonic API documentation, I can provide some insights that may help you better understand Rune's architecture:

1. Rune uses CollectionQuery to retrieve "meta-lists" (lists of playlists) such as listing artists, albums, and playlists. While Subsonic API has methods like getArtists and getPlaylists, Rune implements a unified API wrapper - CollectionQuery - with a list method to retrieve such information. Currently, this API doesn't implement pagination, which seems adequate given the typical requirements for music management.
   I WOULD NOT recommend trying to understand the API design by examining the Dart-Rust backend communication, as the GUI binding implementation could be confusing. It's definitely overcooked, it's my fault.

2. Rune uses Mix Query to get playlists. While Subsonic API has methods like getAlbum, getPlaylist, and getMusicDirectory, we handle these through Mix Query, implemented in the database package with this signature:

pub async fn query_mix_media_files(
    main_db: &DatabaseConnection,
    recommend_db: &RecommendationDbConnection,
    queries: Vec<(String, String)>,
    cursor: usize,
    page_size: usize,
) -> Result<Vec<media_files::Model>> 

In the queries parameter, the tuple's first string describes the operator, and the second is the parameter, for more detail, please check the document from the link above.

3. Rune has two playlist types: Playlist and Mix (similar to iTunes smart playlists). Since Subsonic's API can't manage the latter one, consider implementing them as read-only playlists. They're just playlists with different operators in terms of query encapsulation.

4. For similar song queries (getSimilarSongs API), use MixQuery with a pipe::recommend operator. For example, to find songs similar to track ID 114, use: [("lib::track", "114"), ("pipe::recommend", "-1")].

5. Rune's data structure doesn't support querying "albums by composer" - you may need to implement a separate function for this Subsonic API.

6. For lyrics, use the "parse_audio_lyrics" API in the lyric package, passing the audio file path to get parsed lyrics. It will automatically find the lyric file and parse them to a unified format.

7. Consider creating a separate package for the Subsonic API implementation, with dynamic service enabling/disabling through GlobalParams. For local playback, spawn an async task in initialize_local_player to monitor playback progress via [status_receiver](https://github.com/Losses/rune/blob/9aac026f76d9a881f7b447a9156ba82569d9e00b/native/hub/src/utils/player.rs#L60).

8. Regarding user management, Rune uses a PermissionManager based on device fingerprints rather than username/password.
   Each device generates a TLS key pair, with the public key's SHA256 hash encoded as a 40-character rune string for client access and communication encryption.
   You might find this approach unusual, as the entire authentication system was originally designed to handle remote control functionality - the goal was to enable secure device-to-device communication with minimal keyboard input from users.
   Rather than implementing a separate username/password system, consider using a CRC hash of the rune character for a more user-friendly "password." However, we're open to alternative suggestions.

Feel free to ask me any question if you need any clarification!

[balthild]

Thank you for the clarification!

I'm still trying to set up the project, but I've strugged to bring it up, with the following error:

D:\Projects\Rust\rune on  master is 📦 v1.1.4002 via 🎯 via 🦀 v1.83.0-nightly
❯ flutter run
Connected devices:
Windows (desktop) • windows • windows-x64    • Microsoft Windows [Version 10.0.19045.4842]
Chrome (web)      • chrome  • web-javascript • Google Chrome 131.0.6778.265
[1]: Windows (windows)
[2]: Chrome (chrome)
Please choose one (or "q" to quit): 1
Resolving dependencies...
Downloading packages...
  _fe_analyzer_shared 79.0.0 (80.0.0 available)
  analyzer 7.2.0 (7.3.0 available)
  asn1lib 1.5.8 (1.5.9 available)
> async 2.12.0 (was 2.11.0) (2.13.0 available)
> boolean_selector 2.1.2 (was 2.1.1)
  build_daemon 4.0.3 (4.0.4 available)
  build_resolvers 2.4.3 (2.4.4 available)
  build_runner 2.4.14 (2.4.15 available)
> characters 1.4.0 (was 1.3.0)
> clock 1.1.2 (was 1.1.1)
> collection 1.19.1 (was 1.19.0)
  device_info_plus 11.2.2 (11.3.0 available)
> fake_async 1.3.2 (was 1.3.1) (1.3.3 available)
  ffi 2.1.3 (2.1.4 available)
  file_selector_windows 0.9.3+3 (0.9.3+4 available)
  get 4.6.6 (4.7.2 available)
  get_it 7.7.0 (8.0.3 available)
  intl 0.19.0 (0.20.2 available)
  js 0.7.1 (0.7.2 available)
> leak_tracker 10.0.8 (was 10.0.7) (10.0.9 available)
> leak_tracker_flutter_testing 3.0.9 (was 3.0.8)
> matcher 0.12.17 (was 0.12.16+1)
  material_color_utilities 0.11.1 (0.12.0 available)
  material_symbols_icons 4.2801.1 (4.2805.1 available)
  math_expressions 2.6.0 (2.7.0 available)
> meta 1.16.0 (was 1.15.0)
> path 1.9.1 (was 1.9.0)
  petitparser 6.0.2 (6.1.0 available)
  rinf 7.1.0 (7.1.1 available)
  shelf_web_socket 2.0.1 (3.0.0 available)
> source_span 1.10.1 (was 1.10.0)
> stack_trace 1.12.1 (was 1.12.0)
> stream_channel 2.1.4 (was 2.1.2)
> string_scanner 1.4.1 (was 1.3.0)
> term_glyph 1.2.2 (was 1.2.1)
> test_api 0.7.4 (was 0.7.3)
  uuid 3.0.7 (4.5.1 available)
  vector_graphics 1.1.17 (1.1.18 available)
> vm_service 14.3.1 (was 14.3.0) (15.0.0 available)
  win32 5.10.1 (5.11.0 available)
  win32_registry 1.1.5 (2.0.1 available)
Changed 18 dependencies!
27 packages have newer versions incompatible with dependency constraints.
Try `flutter pub outdated` for more information.
Synthetic package output (package:flutter_gen) is deprecated: https://flutter.dev/to/flutter-gen-deprecation. In a
future release, synthetic-package will default to `false` and will later be removed entirely.
Generating synthetic localizations package failed with 1 error:

Error: The placeholder, count, has its "type" resource attribute set to the "null" type in locale "de", but it is
"Object" in the template placeholder. For compatibility with template placeholder, change the "type" attribute to
"Object".

I searched the error and found that it originates from the i18n toolkit. However, I have no idea how to resolve it.

Update: I tried modifying intl_de.erb and intl_en.erb, adding type field to the count placeholder. Nw flutter complains about another placeholder with same error. I have no idea why flutter recognizes type being Object in intl_en.erb when the field is not exist at all, while in intl_de.erb it recognizes type being null.

[balthild]

1. Rune uses CollectionQuery to retrieve "meta-lists" (lists of playlists) ...

Looks like the API is coupled with the database (sea-orm). Implementing the subsonic support on top of it would be a trouble. Either the CollectionQuery API needs a redesign, or there should be another data abstraction layer between the app and this API (e.g. GraphQL, but I think may not be a right choice for internal communication within one program).

6. For lyrics, use the "parse_audio_lyrics" API in the lyric package, ...

This needs a workaround. The lyrics (as well as the music) are not loaded from the filesystem, and it needs to be get through the getLyricsBySongId API, of which returns a structured lyrics list (JSON) rather than an file needs being parsed. It's more reasonable for me to construct LyricFile directly, rather than generating an LRC and letting parse_audio_lyrics to parse it again.

Regarding user management, Rune uses a PermissionManager ...

Subsonic API authenticates the user with username/password. OpenSubsonic supports key-based authentication, but many implementations don't support it. Also, neither of them could be done without asking for credentials from the user.

[ChardeX03]

Guys I'm willing to pay the paid version of this because it has a potential but may i ask if it going to have a crossfading option for at least 10s and i hope we can have an option when right clicking a song an option to "play selected" and "add to now playing" thanks.

[mncho]

I'm not sure how many people will need this feature, so let's leave it here and see what the community thinks.

We are currently developing a complete and secure remote control protocol and media server for Rune, which I think should solve some requirements, I hope.

Subsonic API would be nice, but a better integrated media server solution which can hopefully be fully open source would be nicer (as subsonic itself is not...) Subsonic API is fairly outdated and personally this one has had no end of issues with it. A problem with the Subsonic API is one of it's supposed strengths which is that it's generic and because of that it has to be a kind of jack of all trades, but it can't fully support any one particular client or have many useful features added on to it.
A purpose-built media server specifically for Rune that integrates fully with it's features and is fully supported by the client will always be preferable to a Subsonic API implementation, so if anything this one thinks that Subsonic support should not be considered until its own media server is done.