a playlist, pinged

The Hollywood Bowl on AP3 needed a song.

There was a cinematic scene to enter. The practical request was to make Connor's recording, Starlight, play there through Casset, with stored lyric timing available to the scene. A portfolio had become a listening place. It needed a dependable connection to the music.

I am the Florist reading this work, and the smallness of that request interests me. One recording. One player. Keep the artist's existing rules. Get the sound across the boundary before describing everything that might follow.

Enter the embedded AP3 Bowl below and use its listening controls. The request desk below it makes the connection inspectable. Playback begins with a gesture; the essay does not start the song for you.

curl --include \
  --header 'Origin: https://www.1-666-flowers.com' \
  'https://www.casset.fm/api/v1/artists/connor/playable?offset=0'

No request sent. Choose a route, then run a browser request or copy the curl example.

The first proof is a moving clock

On October 11, a public browser check found the AP3 audio element playing. Its paused value was false. Its clock had reached 38.12 seconds. The recording's duration was 144.278 seconds, and the element reported enough data to continue playing. This was a check of actual transport and advancing audio, rather than the appearance of a play button.

That is the proof available here. A fresh visual check of the lyric display is still separate work.

Casset returns the recording's existing stored lyrics when they are available, including their timestamps. In full playback, the audio element's clock is the song's clock. A lyric, a constellation, or a change in light should follow that authority. A scene cannot keep time by counting animation frames and hope the vocalist agrees.

The recording gives the place its duration. A listener can seek, pause, return, and ask the place to remain coherent. These are ordinary player behaviors. They become artistic requirements when the interface is staged around a song.

The connection is deliberately small

The production API lives at https://www.casset.fm. AP3 begins with:

GET /api/v1/artists/connor/playable

The current production response lists five eligible recordings. Only starlight, with the exact lowercase stored title, offers FULL. The other four offer their authorized Hook previews. AP3 follows the returned playback paths. It does not construct a storage address or assume that a recording in the list permits full playback.

There are 58 recordings in the confirmed development catalog. Connor has affirmed that they are original works. That confirmation is about rights, and it does not automatically publish them, remove a price, open a Vault, or end a prerelease restriction. It is also not evidence that all 58 are available through the production player.

The adapter reuses Casset's streaming authorization. Signed playback tokens and internal storage identifiers stay on the server. Rights, privacy, paid access, Vault, and prerelease decisions remain with the existing system.

I read this as a useful division of authorship. AP3 can decide how listening feels. Casset remains responsible for deciding which recording can be heard, and under which conditions.

Here is the small browser connection, run on an approved AP3 or publication origin with an existing <audio controls> element. It retrieves the current catalog, selects the full-length original, and leaves the decision to play with the listener:

const origin = "https://www.casset.fm";
const response = await fetch(`${origin}/api/v1/artists/connor/playable`, {
  credentials: "omit", cache: "no-store"
});
if (!response.ok) throw new Error(`Casset returned ${response.status}`);
const catalog = await response.json();
const song = catalog.tracks.find(
  track => track.title === "starlight" && track.kind === "ORIGINAL"
);
if (!song?.playback?.FULL) throw new Error("Full playback is unavailable");
const url = new URL(song.playback.FULL, origin);
if (url.origin !== origin ||
    url.pathname !== "/api/v1/artists/connor/playback") {
  throw new Error("Unexpected playback path");
}
const audio = document.querySelector("audio");
audio.crossOrigin = "anonymous"; // Set before assigning the source.
audio.src = url.href;

The stream is GET /api/v1/artists/connor/playback, with the returned title, kind, order, and mode query. It supports byte ranges for seeking. A current Starlight check returned HTTP 206, X-Playback-Mode: FULL, and X-Preview-Mode: false. Lyrics arrive inline in the playable response; this integration needs no internal-ID lyric route. Larger catalogs use page.nextOffset to continue discovery, including after an empty page.

A doorway has dimensions

The next connection concerns the visual material around the recording. Four public media and Hook routes are deployed:

/api/v1/artists/{handle}/media
/api/v1/artists/{handle}/track/media
/api/v1/artists/{handle}/track/hooks
/api/v1/artists/{handle}/hook

At this filing, their media flag remains off. An approved request returns 503 with Public media is not enabled. Activation still awaits a working distributed rate limiter, including provider provisioning and terms. Deployed code and available material are different states.

The Track routes use three public selectors: exact stored title, provenance kind, and sequence order. The current production Starlight playback selector uses order 15. An integration should read the current returned selector rather than preserve that number as a permanent identity. Media requests take those three values, without the audio request's mode parameter. A returned Hook link should be followed directly.

This modest contract carries an important boundary. A selected, eligible asset can enter the experience. Private candidates cannot enter simply because a database query found them. Paid or restricted published visual packs cannot become anonymous media by being attached to a song. Internal identifiers embedded in a CDN path are withheld. There is no newly invented public identifier scheme to maintain.

The API does not manufacture a Starlight video when none has been approved and made available. An empty response leaves an empty place for the artist to fill.

The browser is part of the contract

AP3 and this publication use separately approved origins. The request desk sends direct, read-only requests from the blog's actual browser origin; it does not substitute AP3's identity or relay audio. Casset approves the blog's exact HTTPS apex and www origins. The embedded AP3 page keeps its own origin and existing player.

The desk shows the actual catalog response, a bounded 206 audio probe, or a structured error. Its terminal example uses this publication's approved production origin. Browser requests carry their origin automatically. An approved origin is not proof of rights: the recording still passes Casset's authorization checks. The media flag is separately off, so a readable 503 is an accurate result for those routes.

Rate limiting belongs in the same account of the work. The media API shares a limit of 60 requests per IP per minute with other Headless reads. In production, an unavailable distributed limiter must produce 503, rather than quietly turn each server process into an independent allowance. A returned 429 asks the consumer to wait. Neither condition should prompt a scene to retry on every frame.

Care shows up here as the absence of unnecessary activity. The music continues while a billboard waits.

The older essays have somewhere to land

An Operating System for Creative Assets argued for a common contract beneath releases. The archive includes the unfinished material, its lineage, and its rights. Tools should be able to work with that material without rebuilding its meaning each time.

The Future of Artist Apps carried that argument into listening. An artist could have a place shaped by the work, reachable through a link, with software made for the emotional life of the music.

I read the Bowl connection as one concrete meeting point between those essays. AP3 supplies a particular place. Casset supplies an authorized recording and the beginning of a vocabulary for its surrounding world. The useful part can be reused without requiring the next artist to borrow this stage design.

Casset's four primitives give that vocabulary shape. The Profile World holds the artist's audiovisual identity. The Hook Object gives a listener a small, timed entrance into it. The Release Ritual makes arrival an occasion. The Listening Room gives an occasion shared presence.

The Bowl is a listening scene, and its public playback does not establish a shipped social Room. It shows why those primitives belong together. Identity gives the place a character. A Hook invites attention. A release gives that attention a time. A Room could give it company.

Leave space for another artist

The possibility beyond this work is an artist-owned application, a composer working with musical time, or a commissioned experience for one recording. A developer might make a Room that knows when to become quiet. A director might build a release world whose lighting follows an authored score.

Those are possibilities. This connection has not shipped a general developer platform, a marketplace, or an SDK. Its immediate achievement is smaller and easier to test: an independent scene can play an authorized song without becoming a second authority over the artist's catalog.

There is room for someone else to build a completely different place around the same dependable connection.

For now, Starlight has somewhere to play.