Flows: playlists that rebuild themselves on a schedule

This is the feature that separates a real discovery server from a generator that hands you a dead list of names. A static playlist tells you about music. A Flow goes and gets it.

Flow files are temporary and live under a _flows folder inside the Downloads Folder. A Flow can publish itself to your playback server without those files ever entering the library, which is what keeps a weekly mix from bloating your permanent collection by thirty tracks a week.

Three groups of settings

Every Flow is configured in three places. Schedule decides when it runs. Source mix decides where the tracks come from. Focus decides what the one hand-configured source looks for.

Schedule

• Tracks sets the playlist size. This is the single number that changes the character of the mix.
• Update days sets which days the Flow may run on.
• Update hour sets the hour of those runs.

A new Flow does nothing until you switch it on. Switching it off keeps every setting so you can pause a Flow over the summer without losing its configuration.

Source mix

Four sources feed a Flow. Only Library uses artists you already own.

One slider sets the share of each enabled source. With four on, it is a four-way mix. With two, a two-way split. With one, that source takes everything, and the numbers on the slider show how many tracks each one actually supplies at the current size.

Turning a source off removes only that source. Your Focus tags, related artists and year range are kept, so switching Focus off and back on does not mean retyping your preferences.

Focus, the one source you configure by hand

Focus does not affect Discover, Library or Trending. You give it three optional inputs and it looks outward for matches.

• Tags set the genres, decades or moods to look for.
• Related artists set the artists it finds similar artists for.
• Release years limit the result to a range. Leave it blank for no limit.

Focus needs at least one tag or one related artist before it can be enabled. Separate multiple entries with commas; duplicates are ignored.

Year ranges are worth understanding precisely. 1980 to 1989 gives you one decade. 2020 to blank gives you 2020 onwards. Blank to 2000 gives you 2000 and earlier. While a range is set, Focus skips tracks whose year is unknown rather than guessing. The other three sources ignore the range entirely, which is a common source of confusion.

Deep Dive

Deep Dive changes which tracks come from each artist. Off, a Flow takes that artist's top ten, which means your most-played artists dominate any mix they appear in. On, it takes tracks ten through twenty-five, so those artists contribute something you have probably not heard. If an artist does not have enough tracks in that range, it falls back to the top ten.

Deep Dive does not change the schedule or the source mix, and it does not apply to any other part of the system. It is purely a Flow-level switch.

How Focus widens its search

When you give Focus several tags and several related artists, it starts strict and relaxes only when it needs more tracks. The order is fixed:

1. Artists related to all your artists, matching all your tags
2. Artists related to all your artists, matching one or more tags
3. Artists related to any of your artists, matching all your tags
4. Artists related to any of your artists, matching one or more tags
5. Artists related to all your artists, with any tags
6. Artists related to any of your artists, with any tags
7. Artists matching all your tags, with no artist constraint
8. Artists matching one or more tags, with no artist constraint

Two tags together therefore find artists matching both first. Only when that produces too few candidates does it relax to one or the other. If you want a Flow that follows Focus closely, give Focus a high share or run it alone.

What happens when a Flow runs

1. Track count and each source's share are set.
2. A large candidate set is collected for every enabled source, and a separate set for Focus when it is on.
3. A playlist is picked with one song per artist.
4. Any shortfall in one source transfers to the other sources.
5. A reserve of candidates from the same run is held back.
6. The playlist goes to the download worker.
7. If a download fails, the track is replaced from the reserve. Only Flows do this.

That reserve is the difference between a Flow that keeps its shape and a playlist that comes back three tracks short every week.

Scrobbling and the library

Two per-Flow switches decide how much the Flow disturbs the rest of the system.

• Scrobble tracks. On, plays from this Flow go into your listening history and therefore influence future recommendations and your scrobbling services. Off, they do not. Plays recorded before you flipped it stay recorded.
• Show in library.Off by default. On, the Flow's current tracks appear on the library pages, and leave when they rotate out, when you turn it off, or when you delete the Flow. Aurral does not move or copy any files, and neither Lidarr nor your playback servers change.

Templates

A new Flow can start blank or from a template.

• Discover Weekly picks from your recommendations.
• Trending Mix picks from trending artists.
• Library Blend picks from artists in your library.
• Listening History focuses on artists you have played recently. Only appears once Aurral has history for you.
• Release Radar takes one track from each recent album you do not have yet. Its recipe is fixed; you can only change its size and schedule.

Every template except Release Radar can be edited afterwards, so the starting point is a convenience rather than a commitment.

Keeping, replacing and exporting

One Flow track can be promoted to the library from its track menu. To save everything a Flow has finished as a fixed playlist, use Save as playlist in the Flow menu, which creates a static playlist named after the Flow with Static appended.

To replace a finished track, right-click it and choose Re-search. Automatic lets Aurral pick a new recording; Manual opens a search in one download client so you can choose the result yourself. The current file stays available until the replacement passes every check.

Flow tracks also live in a Lidarr-visible feed. Each Flow can publish a URL that Lidarr reads as a custom list, which is a tidy way to keep the two tools pointed at the same list.