Commit graph sonda/internal
Author SHA1 Message Date
tgiachi
14710c3a55 Merge remote-tracking branch 'origin/main' into feat/show-version 2026-11-15 17:31:25 +01:00
tgiachi
a880b87a9e chore: move the module path to git.kmalloc.org/squid/sonda
Migrating hosting from Codeberg to git.kmalloc.org. Renames the
module path in go.mod, every internal import, and the OCI image
source label in the Dockerfile. Build and full test suite verified
green after the rename.
2026-11-15 17:30:29 +01:00
tgiachi
00e5254fde feat: show the version of Sonda
- GET /version (signed-in users only) answers the version and, when the binary knows it, the commit it was built from and whether the files had uncommitted changes
- the commit is read from the build info Go records in a git checkout, so no build flag is needed
- the side menu shows the version under the privacy line, and Settings has an About section with it
- the menu's own requests (friends, version) are not counted as requests of an admin page in the Shell tests
2026-11-15 17:30:14 +01:00
tgiachi
af933ad9ea feat(lidarr): a page of the artists I sent to Lidarr, and whether they arrived
Sonda kept no record of what an admin sent to Lidarr. Each send is now kept
(migration 0033, lidarr_requests), once for each artist, with the first date.
GET /lidarr/requests lists them newest first and says for each whether its
music is in the library: it is noticed when the list is read, by the artist
that was sent or by an artist of the same name that a scan made, and the day
it first was is kept.

The web app has a Requests page for admins, under Discover in the menu and on
Home on a phone, with the waiting and the arrived artists apart.
2026-11-15 12:32:29 +01:00
tgiachi
9f63121028 feat(history): browse my listening by year and month
GET /listens/months counts the listens that count in each calendar month,
newest first, in the time zone the client sends (the zone database is built
into the binary). GET /listens takes from and until to read one month, and
each listen now names its artists.

The web app has a History page: a button for each year with its listens, the
months of the year as bars, and the listens of the chosen month a day at a
time, with the choice kept in the address.
2026-11-15 12:07:19 +01:00
tgiachi
7c961028bb feat(discover): suggestions learned from my own listening
A new model (internal/affinity) splits my listening into sittings (plays less
than 30 minutes apart) and links the artists that are often played in the same
sitting. It is built from my listening events and from the plays of artists I
have no music of that an import kept, on my server, with no provider, and is
built again when plays are added (migration 0032 keeps it as a cache).

GET /me/suggestions gives two lists, each artist with the artists it goes
with: the ones I have and did not play for two months, and the ones I played
at least three times and have no music of. It needs discover_from_listening.
POST /me/suggestions/wishlist puts an artist I played on my wishlist by name.
Discover has a "From your listening" tab for both lists.
2026-11-15 11:09:19 +01:00
tgiachi
3ba087bcc5 feat(import): keep the artist and the time of plays of artists I do not have
The plays of songs that are not in the library were only counted. Now, when
such a play counts as listened (not a skip, long enough), its artist and its
time are kept in a new table (migration 0031, outside_plays), so a model can
later learn which artists are played together, also those not in the library.
No title, album, IP address, country or platform is stored. The same play is
kept once however many times the export is imported, and the report says how
many were kept (history.kept_outside).
2026-11-15 10:50:29 +01:00
tgiachi
aa4883b955 feat(import): an Import page, with the numbers of every import kept
The report of an import used to be shown once in Settings and lost. Each
import is now stored (migration 0030, spotify_imports) as the report in
JSON, and GET /me/import/spotify lists the past ones, newest first.
POST /me/import/spotify answers with the same shape plus id and created_at.

The web app gets an Import page (menu, More tab) with the form and one card
for each past import: plays read, added, already there, not in the library,
the playlists it made and the artists that are missing. Settings only links
to it. An import that finds nothing is not kept, and a report that cannot be
stored is still returned to the person who asked.
2026-11-15 10:34:12 +01:00
tgiachi
c08a234159 feat(import): bring my Spotify history and playlists in
I want my listening history and my playlists from Spotify in Sonda, so that Mixes, Discover and the smart playlists know what I listened to before.

- POST /me/import/spotify takes the zip of a Spotify data export, or the JSON files in it (extended streaming history, the older StreamingHistory files, the playlists), as a multipart form. A request may be up to 512 MB; the zip is limited in how much it may inflate and how many files it may hold, and only the history and playlist files are opened.
- Everything is read on the server and nothing is sent anywhere. The files are not kept, and of each play only the song, artist, album, time and how it ended are read: never the IP addresses, the country or the platform.
- A play is matched to a song of the library by artist and title (no case, accents or punctuation; a "Remastered" note on either side; any credited artist; a compilation by its album). A match is recorded as a listen at the time of the play: complete, skip or play. Nothing is posted to Mastodon.
- Importing the same files again records nothing twice, and the same play in the two shapes of history file is counted once. One import per person at a time.
- A playlist becomes a playlist of mine with the songs that were found; one with a name I already have is left alone, and one with none of its songs found is not made.
- The report counts what was added, what was already there and what is not in the library, and lists the missing artists, most played first.
- Settings gets an Import from Spotify section; the API client can send files.
- English and Italian.
2026-11-14 16:25:58 +01:00
tgiachi
a92326c1b4 feat(player): play every song at a similar volume with ReplayGain
I want songs from different albums to play at about the same loudness without touching the volume.

- The scan reads REPLAYGAIN_TRACK_GAIN, REPLAYGAIN_ALBUM_GAIN and the two peaks from each file, offline, and keeps them on the file (migration 0029, SQLite and PostgreSQL). A tag that makes no sense (not a number, more than 60 dB) is left out. Files indexed before have none until a full rescan.
- A track in the API carries them as replay_gain (null when its first playable file says nothing).
- The web player levels each song with them: the sound passes through a gain node, so a quiet song can be raised too, but never so much that its peak would clip (and never more than 12 dB when there is no peak). Without Web Audio the element is turned down instead.
- A new Playback section in Settings chooses song by song (default), album by album or off; the choice stays on the device.
- README: the ReplayGain entry, and the old line that said the year was not in the index is removed.
- English and Italian.
2026-11-14 16:02:23 +01:00
tgiachi
246cdc63d3 feat: Recently added on Home, with a newest-first order for albums
I can now see on Home what I added to the library lately.

- GET /albums takes sort=recent: the albums added last first, by the date of the newest file of the album that is not missing, then the newest album. The default stays by title and any other value is invalid_input.
- The albums shelf on Home is Recently added and asks for that order; See all leads to /albums?sort=recent, which the albums page understands.
- English and Italian.
2026-11-14 10:29:08 +01:00
tgiachi
d3695d00d6 feat(smart): smart playlists by the year a song came out
I can now ask for songs from a year to a year, for example the eighties.

- The year is read from the tags of each file (the original date first, then the date) and kept on the track: migration 0028 adds tracks.year (SQLite and PostgreSQL).
- Two new rules, min_year and max_year, both included and checked (1000 to 2999, the first not after the second). The index answers them. A song with no year never matches.
- Files indexed before have no year until a full rescan reads their tags again; the editor and the API description say so.
- The smart playlist editor gets From year and Up to year, and the rules are described in words.
- English and Italian.
2026-11-14 10:19:39 +01:00
tgiachi
a7a258464b feat(mastodon): say in Settings when the instance no longer accepts the token
When the instance refuses the token on a post (it was revoked or expired), the account now remembers it: GET /me/mastodon says problem: token_refused and nothing is posted until I connect again with a new token.

- migration 0027 adds mastodon_accounts.problem (SQLite and PostgreSQL).
- Only a refused token marks the account; an instance that is down does not.
- Saving the account again clears the problem.
- Settings shows the warning and the form with the instance already filled in, next to Disconnect.
- English and Italian.
2026-11-14 10:04:07 +01:00
tgiachi
0dc7782028 feat: post what I listen to on Mastodon, if I connect an account
In Settings a person connects their Mastodon account: the address of their instance and an access token that can only write posts. Nothing is connected by default, and disconnecting forgets the token.

- When a track of this server counts as a listen, it is posted as an unlisted status: title, artists, album and #NowPlaying, with no link. Music of a friend is never reported as a listen, so it is never posted. At most one post a minute for each person, in the background, with a bounded queue.
- PUT, GET and DELETE /me/mastodon. The token is never sent back.
- The address of the instance is typed by a person, so it must be a domain name on https, and the client never connects to a private, loopback or link-local address (checked on the address it really dials, so a name that points inside is no way in), never follows a redirect, ignores proxy settings and reads at most a megabyte.
- New table mastodon_accounts (SQLite and PostgreSQL) that goes with the account.
- README: the feature, and the credit for butterchurn and its presets.
2026-11-13 21:59:03 +01:00
tgiachi
7243fe7a04 feat: the seek bar shows the waveform of a friend's track
A friend's server works out the peaks of a track in a shared library and gives
them to a signed friend (GET /peer/tracks/{id}/peaks). This server asks for them
and passes them on to the listener after checking them: 1 to 1000 levels from 0
to 255, as many as the count says (GET /friends/{friend}/tracks/{id}/peaks). The
player uses it for a friend's track, so its bar is no longer flat.
2026-11-13 19:46:37 +01:00
tgiachi
bfa3e6e458 feat: the seek bar shows the waveform of the track
The seek bar of the player is the shape of the track, with the part already
heard in colour, as in the approved mockups.

- GET /tracks/{id}/peaks gives 120 levels from 0 to 255 (the loudest part is
  255 whatever the volume). The server finds them by decoding the file, MP3 with
  go-mp3 and FLAC with eaburns/flac, both in pure Go. It takes a few seconds, so
  the first request starts the work in the background and answers 202 with
  Retry-After. It is done once for each file (migration 0025), worked out again
  if the size or the date of the file change, one track at a time so that a NAS
  is not kept busy, and a file that cannot be decoded is remembered as such. A
  track deleted while its file is being decoded is not an error.
- The seek bar of the player bar and of the full-screen player draws the shape
  and asks again every few seconds until it is there. Until then, and for a
  friend's track, which has none, the bar is flat. The heard part is teal, or
  the friend colour. It is still a native range input, 44 pixels high in the
  bar.
- OpenAPI path and schema, contract test, tests with real MP3 and FLAC files,
  and a real server checked: a 4-minute MP3 is ready in about a second and a
  half, a 12 MB FLAC in three, and asking again takes a millisecond.
2026-11-13 17:35:00 +01:00
tgiachi
7f38ae038e feat(federation): play a friend's music and show its covers through this server
A listener can now play the music of a friend from the web app.

- GET /friends/{friend}/tracks/{id}/stream and /friends/{friend}/albums/{id}/cover,
  in the group that accepts the access token in the query, like the local ones.
  This server asks the friend with a signed request and passes the bytes on,
  forwarding Range and the conditional headers, so a player can seek. The
  friend sees this server and never the listener.
- Only audio (for a stream) and a JPEG (for a cover) are passed on, with a
  fixed set of headers: Content-Type, Content-Length, Content-Range,
  Accept-Ranges, Last-Modified and ETag. Anything else a friend sends, or
  what a listener's request carries (cookies, authorization), is not passed
  either way, so a friend cannot make this server serve anything else.
- The client of friends can open a streaming answer: no limit on its whole
  length, one only on its beginning, no redirect followed, and it ends when
  the listener leaves.
- The player holds a friend's track next to the local ones. It streams
  through this server, shows the friend's cover, links to the music of the
  friend and not to a local page, is told apart from a local track with the
  same id, and is never reported as a listen: it does not feed taste.
- The songs of a friend have play buttons, the albums their covers.
- OpenAPI paths, contract test, and two real servers checked end to end: the
  bytes of the file arrive unchanged, with ranges, and a song plays in a
  browser from the friend's server through the other one.
2026-11-13 16:11:59 +01:00
tgiachi
29bd46c3ba feat(federation): browse and search a friend's catalogue
A listener can now browse the music that a friend's server shares.

- GET /friends lists the friends both admins agreed to, by id and name; an
  address or a key is for the admin only. It is empty while federation is off.
- GET /friends/{friend}/artists, albums and tracks (q, limit, offset,
  artist_id, album_id) and the one-item routes: this server asks the friend
  with a request signed by its own key and passes the page on. The answer is
  read into this API's own types first, so what a friend adds is dropped and
  the shape is the one in the spec. Nothing is kept, and the ids are the
  friend's own.
- Only the three kinds of thing can be asked, and only the parameters their
  routes take: a listener cannot make this server ask a friend for anything
  else. The friend's address and what the connection said never reach a
  listener.
- A friend that cannot be reached, no longer takes this server for a friend,
  is busy or is not there is told in words, never as an error page.
- The client of friends gets a generic signed GET with a size limit.
- The web app has a Friends entry in the menu, shown only when there is a
  friend to browse, a list of friends and a page for each with artists,
  albums and songs, a search and links from an artist to its albums and from
  an album to its songs, in English and Italian.
- OpenAPI paths and schemas, contract test with a friend that is a whole
  server, and two real servers checked end to end.
2026-11-13 15:59:26 +01:00
tgiachi
89826fcd07 feat(federation): shared libraries and the signed read API for friends
A friend's server can now read what an admin chose to share.

- A library has a shared flag, off by default (migration 0024), set with
  PATCH /libraries/{id}. Changing it never waits for a scan.
- Requests between friends are signed with each server's Ed25519 key: the
  method, the path and query from /api/v1/peer on, a date, a nonce and the
  hash of the empty body. A date more than two minutes off, a nonce used
  before, an unknown or removed key and a bad signature all get the same
  401. A friend still waiting for approval gets 403, and while federation is
  off the routes are missing.
- Routes under /api/v1/peer: info, artists, albums, tracks, covers and
  streaming with Range. The catalogue, the counts and the format come from
  the shared libraries only, and the file served is the one in a shared
  library. No path, library id, listen, taste, user or provider data is part
  of an answer.
- A friend never makes this server build a cover, read an audio file or ask
  a provider: only covers already built from the audio files are served.
- Per-friend limits, set by the admin: parallel streams (default 3) and KiB
  a second (default 4096, 0 for none), shared by all the streams of a friend.
- A log of the last requests friends made (route, status, bytes; never the
  query), kept for 7 days, in GET /federation/requests.
- POST /federation/friends/{id}/check lets the side that pasted an invite
  find out that the other admin approved; a signed call from a friend this
  server is waiting for makes it active too.
- Admin: a Shared with friends switch on each library, which asks first, and
  limits, recent requests and a Check button in the Federation section, in
  English and Italian.
- OpenAPI paths and schemas, contract test, two real servers checked end to
  end with tagged audio files.
2026-11-13 15:04:07 +01:00
tgiachi
695a012f6c feat(federation): keys, invites and pairing between Sonda servers
Two servers can become friends. Federation is off by default; turning it
on makes an Ed25519 key pair, kept in the database and never returned.

- Admin routes: GET and PUT /federation (on or off, name, public address),
  POST /federation/invites, POST /federation/friends (use an invite),
  POST /federation/friends/{id}/approve and DELETE /federation/friends/{id}.
- POST /api/v1/peer/pair, the one route without a login: the code of an
  invite is the proof. An invite works once and expires in 24 hours; only
  a hash of its code is stored. At most 10 attempts a minute.
- The server that uses an invite checks that the key it gets back is the
  one named in the invite. A friend is kept as incoming or outgoing until
  both admins have agreed.
- https only, except plain http on a network of one's own (private,
  loopback or link-local address, or a local name).
- A client of its own for friends, with no redirects, no proxy from the
  environment, a size limit and a time limit; the providers' client stays
  as it is.
- Migration 0023 for SQLite and PostgreSQL, OpenAPI paths and schemas,
  contract test with two real servers.
- Admin > Federation in the web app, in English and Italian, and a
  section in the README.
2026-11-13 14:43:48 +01:00
tgiachi
2217a12042 feat(web): the app can be installed and opens offline (PWA)
- A web app manifest with the name, the colours, the display mode and the icons (also a maskable one), so a browser offers to install it; Settings has an Install button where the browser lets the page ask.
- A service worker keeps the files of the app (the page, its scripts and styles, the manifest and the icons) so it opens at once, and shows its own page when the server cannot be reached. It never touches /api/: nothing from the API is kept, answered from a cache, or asked for by the worker. Files that failed, and anything that is not a file of the app, are not kept; the scripts it keeps are capped.
- A new version waits until the page says it may take over: a line says a new version is ready and its button reloads the page after the new worker took over.
- The server gives the manifest its type (application/manifest+json); the worker and the manifest are asked again every time, and the policy of the page needs no change.
2026-11-13 12:22:00 +01:00
tgiachi
82b92eaf08 feat(smart): covers on the Smart playlists page
- GET /smart-playlists/{id}/covers and GET /smart-playlists/suggested/{key}/covers say which albums stand for a smart playlist: up to 4 albums that have a cover, in the order of the songs the rules give now (for the variant asked), each once. They are worked out apart from the songs, and the answer may be kept for a few minutes.
- The cards of the Smart playlists page are a grid with the collage. The list shows at once and each card asks for its own covers after that, showing the rings until they arrive or when they cannot be read.
- The pages of a smart playlist and of a suggested one use the same covers for the variant on screen, so only albums that have a cover are shown.
2026-11-13 11:41:23 +01:00
tgiachi
a1e192066c feat: playlists with a cover, a collage of the covers of their albums
- A playlist says which albums stand for it (album_ids): up to 4 albums that have a cover, in the order they first appear in it, each once, read for a whole page of playlists in one query. An empty one has an empty list.
- The Playlists page is a grid of cards with the collage, and the page of a playlist shows it in its header. The pages of a smart playlist and of a suggested one show a collage of the albums of their first songs.
- Nothing is fetched for it: the covers are the ones the library already has.
2026-11-13 11:32:01 +01:00
tgiachi
1b79a7370d feat(naming): an admin section that shows what Claude costs
- Every question asked of Claude is recorded: when, the kind of list, the model, the tokens read and written (also for an answer that was no name but was billed) and how it ended: a name made, a name taken from the kept ones, or a failure and why. Nothing of the facts that were sent is recorded. Records older than a year are removed (migration 0022).
- GET /providers/anthropic/usage (admin) gives the totals of the last days (1 to 365, 30 by default), a row for each day with calls and an estimate of the cost in dollars, worked out from the tokens and the price of the model. It says it is an estimate and flags calls to a model whose price is not known.
- A Costs section in the admin area shows the last 7, 30 or 90 days.
2026-11-13 11:14:45 +01:00
tgiachi
0d8fadadf6 feat(naming): playlists named by Claude by themselves when it is on
- Keeping a mix, a smart playlist or a suggested smart playlist as a playlist with auto_name gives it the name and the line of description that Claude makes from its songs. The name sent stays the fallback when the provider is off, or Claude fails or declines, so a failure never stops a playlist from being made.
- A smart playlist can be made with no name when Claude is on: it names it from the songs the rules give. Without Claude a name is still needed.
- GET /naming says whether names are automatic, so the editor does not ask for a name that Claude will make.
- The web app always asks for an automatic name and sends the language of the page.
2026-11-13 11:00:23 +01:00
tgiachi
bf5c108785 feat(smart): the songs I play most, in rotation lately and over the year
- A smart playlist rule, plays_within_days, that counts only the plays of the last N days where plays are counted: the minimum of plays and the most and least played orders. The other listening rules are not changed.
- Two more suggested smart playlists on it: on-rotation (the most played of the last 30 days, at least twice) and top-year (the same over 365 days). most-played stays as the plays of all time.
- The editor has the new rule and the page says it in words. Texts in English and Italian.
2026-11-13 10:44:55 +01:00
tgiachi
23ae396430 feat(naming): names for playlists, with the Claude API as an option
- A pipeline over the facts about a list of songs (up to 5 genres, up to 8 artist names, how many songs, a few words about the rules of a smart playlist, the language). A local namer builds a name from the genres and the artists on this server and is always there; Claude is asked first only when an admin turned the new anthropic provider on with an API key. A failure or a refusal gives the local name. What Claude answers is kept under a hash of the facts, so the same question is not asked or paid twice.
- The anthropic provider: a write-only key (kept as typed, it is case-sensitive), a test that reads one model and sends no text, and a plain statement of what is sent. It uses the official Go SDK with nothing taken from the environment of the server (no key, profile, base URL or extra headers), no redirects, no system headers, and an answer that is cleaned and bounded.
- POST /playlists/{id}/suggest-name, POST /mixes/{id}/suggest-name and POST /smart-playlists/{id}/suggest-name answer with a name, a description and where it came from.
2026-11-13 10:33:48 +01:00
tgiachi
17f70d582d feat(smart): smart playlists made for everybody, without making them
- A fixed set that is there with no setup: recently added, never played, most played, not heard for a while, high quality files and long tracks. The rules are the same for everyone and are worked out on each one's own library and history, like the mixes; nothing is stored, so they follow the library.
- GET /smart-playlists/suggested, GET /smart-playlists/suggested/{key}/tracks and POST /smart-playlists/suggested/{key}/playlist.
- The Smart playlists page begins with them, with names and descriptions in the language of the page. A suggested one can be played, shuffled again, kept as a playlist, or made one's own: a copy of its rules opens in the editor.
2026-11-13 10:06:58 +01:00
tgiachi
1ad0d718c8 feat(mixes): a picture for each mix, a collage of album covers
- The server says which albums stand for a mix (album_ids): up to 4 albums that have a cover, by different artists, with most of the songs the mix draws from. The list gives them for every mix and a mix gives its own.
- The Mixes page shows the collage on each card (a 2 by 2 collage with four covers, the first one alone with fewer, the rings with none), and the page of a mix shows it in its header.
- Nothing is fetched for it: a mix has a picture once the covers of its albums were looked up.
2026-11-13 09:58:04 +01:00
tgiachi
a607070149 feat(smart): GET /genres, the genres of the library to choose from
Each genre with the number of playable songs, the biggest first and by name among equals. The spellings the files have are one genre, written the way the rules compare them, and tags such as other or unknown are left out.
2026-11-13 09:46:03 +01:00
tgiachi
104e41d10a feat(smart): smart playlists, playlists made of rules
A smart playlist is a set of rules, an order and a limit, kept under a name and worked out again every time it is opened.

- Rules (all that are set must hold): genres, artists and albums (is, or is not), file formats and a minimum bitrate of the same file, added in the last N days, never played, played or not played in the last N days, played at least N times, skipped or not skipped lately, only liked artists or without the avoided ones, and the length of the song.
- Orders: random (the same all day, a variant gives another), most played, least played, recently added, by title. A limit of 50 by default, 500 at most.
- GET/POST /smart-playlists, GET/PATCH/DELETE /smart-playlists/{id}, GET /smart-playlists/{id}/tracks, POST /smart-playlists/preview to try rules without keeping them, and POST /smart-playlists/{id}/playlist to keep what is shown as a normal playlist.
- Only the rules are stored (migration 0020); the songs come from the files and from the owner's own history and taste. Nothing is sent to a provider, and a smart playlist is its owner's alone. A user can have 200.
- Mixes and smart playlists share the way a list of songs becomes a playlist.
2026-11-13 09:36:52 +01:00
tgiachi
25f41dd5fc test(mixes): remove a throwaway check against a local database
It was committed by mistake with the mixes: it reads a database from a path on one machine and prints, so it checks nothing.
2026-11-13 09:23:11 +01:00
tgiachi
6b40ea06fa fix(musicbrainz): follow the redirect of an id that was merged
When MusicBrainz merges two entries, the old id answers with a redirect to the new one. The client refused every redirect, so the description of that artist failed at every run of the job.

The client for MusicBrainz now follows a redirect only to https://musicbrainz.org/ws/2/artist/<id> or /release-group/<id>, with no credentials or port. Any other redirect is still refused. The new id is used for that request only.
2026-11-12 19:34:59 +01:00
tgiachi
94dbf57544 feat(mixes): playlists made for a listener, by listening and by genre
GET /mixes lists the mixes a listener has, GET /mixes/{id}?variant= gives one with its
songs, and POST /mixes/{id}/playlist keeps one as a playlist of their own.

- Genre mixes: one for each genre with at least 30 songs by 4 artists, the genres the
  listener plays most first. Spellings that differ by case or punctuation are one
  genre (jazz+funk, jazz-funk); other and unknown are not genres.
- Made for the listener: Your favourites, Rediscover (liked artists not played for a
  month; both need history) and New to you (songs never played).
- A mix has up to 50 songs, at most 3 by one artist, never two of an artist in a row,
  picked by weighted chance (artist preference, never-played, less if played lately;
  skipped lately and avoided artists are out). The seed is user, mix, day and variant,
  so a mix is steady for a day and another is one variant away.
- Worked out when asked, from the listener's own history and the genre tags; nothing is
  stored and nothing leaves the server.

internal/mixes holds the selection and the service, internal/storage/mixes.go the
queries (SQLite and PostgreSQL). Spec, contract test, README and web types follow;
the web app comes separately.

Closes #95
2026-11-12 17:25:49 +01:00
tgiachi
ce109a14a3 fix(duplicates): move the playlist items of a removed copy to the one that stays
playlist_items.track_id is ON DELETE CASCADE (added by the playlists work on main), so deleting a
duplicate would have taken the song out of every user's playlists, as it would have done with the
listens. RemoveCopy now repoints the items first, in the same transaction.

Also merges main (playlists) into the branch.
2026-11-12 17:01:36 +01:00
tgiachi
3d986f051d Merge remote-tracking branch 'origin/main' into feat/duplicates-api 2026-11-12 17:00:00 +01:00
tgiachi
f900a22904 feat(playlists): a user keeps playlists of tracks
Each user can make playlists: a name (1 to 200 characters), an optional
description and an ordered list of tracks, where a track may appear more than once
(each place has its own item id). The API makes, lists, renames and deletes
playlists, and adds tracks (at the end or at a position), lists them, moves one and
removes one; positions count from 0 without gaps.

A playlist is its owner's alone: somebody else's answers playlist_not_found (404),
as one that does not exist. A track whose files were deleted or moved stays in the
playlist and is listed with available false. Deleting a user deletes their
playlists. Limits: 1,000 playlists per user, 10,000 tracks per playlist, 500 tracks
per request.

Changes of one playlist are serialised by locking its row, so two requests at once
keep the positions unique on PostgreSQL too. Migration 0019, OpenAPI spec, contract
test, README and the web API types follow. The web app comes separately.

Closes #89
2026-11-12 16:55:25 +01:00
tgiachi
fff24dfe25 Merge main into feat/duplicates-api 2026-11-12 16:53:50 +01:00
tgiachi
05f4e33470 fix(duplicates): finish updating the index even if the client leaves after the file is deleted
Found reviewing the branch: a request cancelled between deleting the file and updating the index left a track whose file does not exist. The index step no longer depends on the request's context. Test first: it cancels the request at exactly that moment.
2026-11-12 16:52:38 +01:00
tgiachi
30f45bb383 feat(api): list the duplicate songs, check their files and delete the extra copies
Admin only. GET /library/duplicates lists the groups with what is needed to choose and never a path; POST|GET|DELETE /library/duplicates/check is the job that fingerprints the files; POST /library/duplicates/clean deletes the chosen copies of one song, each refusal with its own status and code (deletion_disabled, scan_running, invalid_input, track_not_found, duplicate_changed, library_read_only). In the spec with the contract test, and wired in main, where an info line says when library.allow_delete is on.
2026-11-12 16:46:50 +01:00
tgiachi
d3e3bde2e4 feat(duplicates): find the copies, fingerprint their files and remove the extra ones
Groups come from the index. The check job reads the files of the groups and keeps their SHA-256, skipping a file that changed since the scan. Clean refuses unless deleting is on and no scan runs, works out again that the tracks are copies of one song, checks every file is still what the scan saw (so the last copy is never removed, and a link or a path that leaves the library is refused) before removing any, removes each file through an os.Root, then takes its track out of the index and moves its listens. Tested with real files, including the refusals; three of the safety rules were broken on purpose to see the tests fail.
2026-11-12 16:40:43 +01:00
tgiachi
65edfc4918 feat(storage): candidates for duplicates, file fingerprints and removing a copy from the index
Candidates are the playable files whose title another file shares. A fingerprint is kept only for the file that was read (same size and date), and a scan that finds the file changed forgets it. Removing a copy moves its listens to the track that stays, because they would be deleted with the track, then removes the track with its file, credits and genres. Tested on SQLite and PostgreSQL.
2026-11-12 16:37:15 +01:00
tgiachi
2669f6c6dc feat(images): ask again without the disc or edition note of a title
Many titles end with a note taken from the tags: (disc 2), (bonus disc),
(Deluxe Edition), sometimes cut off by a tag that is too short. A release group is
named after the album, not after a disc, so those albums found nothing.

The lookup now asks for the title as it is first and, only when MusicBrainz has no
sure result, once more without the notes at its end (baseTitle). A group in brackets
with none of the words of a note stays part of the title, and a title that would be
left empty is kept. The rule for accepting a result does not change.

On 25 albums of a real library with such notes, 19 found a cover (none before).

Closes #87
2026-11-12 16:35:23 +01:00
tgiachi
7e1e63bdd2 feat(duplicates): group the copies of a song and propose the one to keep
The same song is the same set of main artists, the same title and durations within 3 seconds, every pair within it; a file with no duration or no artist never matches. A group says whether its files were all read and are byte for byte the same. The copy to keep is the lossless one, then the higher bitrate, then the bigger file, then the oldest, whatever the order of the input.
2026-11-12 16:34:12 +01:00
tgiachi
309160951e feat(config): library.allow_delete, off by default, for deleting duplicate files
The one setting that lets Sonda delete files from the music folders. Off until the operator turns it on, in the file or with SONDA_ALLOW_DELETE; a value that is not true or false is an error, not ignored.
2026-11-12 16:33:11 +01:00
tgiachi
12f487f223 feat(images): find the cover of an album that has no MusicBrainz id
The covers asked the Cover Art Archive only for albums whose files carried a
release group id. In a library tagged without MusicBrainz, 3,321 of 3,433 albums
have none, so almost no cover could come from the archive.

An album with no id is now looked up in MusicBrainz by the name of its artist and
its title (musicbrainz.Client.FindReleaseGroup), when the musicbrainz provider is
on too. A result is taken only if MusicBrainz scores it 95 or more and its title
and one name of its artist credit equal the ones asked for after folding case,
accents and punctuation; the first such result wins and anything doubtful is no
cover. The id is used for that request and is not stored on the album.

Artists are still never looked up by name. The MusicBrainz and Cover Art Archive
provider texts, the README and the API spec say what is sent.

On 47 albums sampled from a real library, 23 got a cover (all checked as the right
ones by eye or by exact title and artist); a title like 'Live (disc 2)' finds nothing.

Closes #83
2026-11-12 16:22:02 +01:00
tgiachi
249b2cb7bc feat: embed the built web app in sondad behind the webui build tag
internal/webui carries web/'s build (written to internal/webui/dist by Vite, git-ignored) when sondad is built with -tags webui. Without the tag the app is absent and a plain go build and go test still work with no Node. main wires it into the router.
2026-11-12 15:56:34 +01:00
tgiachi
7acac73dd0 feat(api): serve the web app on / and for its own links, with a strict policy
A GET or HEAD outside /api is a file of the app, or its page for a link inside it (/admin/users). A file that is missing (under assets/ or with an extension) answers 404, never the page, so a missing script is not run as HTML; hidden files and folders are never served; everything else, and any unknown /api address, keeps the JSON 404. The page gets a Content-Security-Policy that allows only its own server (no address, no wildcard, no inline script), nosniff, DENY and no referrer. Hashed assets are cached for good and the page is asked again. With no app in the binary, / says how to build it. The app is an optional dependency, so existing wiring is unchanged.
2026-11-12 15:56:33 +01:00
tgiachi
96ad8b1f62 feat(auth): accounts are identified by an email address
Registration, login and the admin's create-user call take `email` instead of
`username`, and the user object returns `email`. The address is checked for its
shape (a plain address, no display name, at most 254 bytes), trimmed, lowercased
and unique. Sonda sends no mail and does not verify it.

- Migration 0018 renames the users column (and the PostgreSQL index); values are
  kept, so an account made before keeps its plain name and still logs in with it
  (login does not check the shape).
- Error code username_taken becomes email_taken; the JWT claim is now `email`,
  so access tokens issued before the upgrade are refused once and clients refresh.
- OpenAPI spec, README, the web login form (type email), strings (en, it) and
  the generated API types follow.

Closes #65
2026-11-12 14:40:33 +01:00
tgiachi
77be31ea4a perf(storage): fix stale match and search keys in batches on PostgreSQL
The refresh at start fixed each stale row with its own UPDATE. On PostgreSQL that
is one round trip per row: 100,000 stale rows took 5.3 s on localhost, and more
over a network. The rows now go 500 at a time in one UPDATE ... FROM (VALUES ...):
1.9 s for the same rows. SQLite keeps one UPDATE per row, which is faster there
(no round trip to save). A run with nothing to fix is unchanged (about 0.1 s per
100,000 rows).

A test with 1,250 stale rows covers more than two batches on both databases.
2026-11-12 14:08:40 +01:00