A boolean set at upload time decides what your video migration costs

Call `audit(mediaId)` and one request goes out: a GET to `/on-demand/{mediaId}` on `https://api.fastpix.com/v1`, with HTTP Basic auth where the Access Token ID is the username and the Secret Key is the password ^[8][9]. The response's `data` comes back as an object ^[9]. A code comment notes that older FastPix docs showed an array there, so a parser written against those breaks on this call ^[9].

The adapter reads `data.playbackIds?.[0]?.id`, builds `https://stream.fastpix.com/{id}.m3u8` from it, and sets `standardHls: Boolean(playbackUrl.endsWith('.m3u8'))` ^[10][11]. It is testing a suffix it appended itself two lines earlier ^[10][11]. The column answers whether a playback ID exists, and says nothing about whether the provider serves a standard manifest ^[12]. So it is true for every asset that has a playback ID and false only when the URL came back empty ^[12].

`sourceRetained` is `data.sourceAccess ?? null`, and `sourceAccess` is decided at ingest on the create call; if it was not enabled then, the original is not retrievable now ^[13]. Because the typedef declares `boolean|null`, three values can reach the CSV: true, false, and null where the field was absent ^[7][14]. A plan that reads null as false will under-count the assets it has to re-encode ^[14].

The post recommends running the script once a quarter ^[1]. Four runs a year is one every 91 days ^[19]. Since retention is fixed at upload, that interval is also the largest amount of ingest that can arrive with the wrong default before a row says so ^[19]. "If it defaulted off across your whole library, this script will tell you today rather than during a migration in 2028," the post writes ^[16]. Today is the right answer for a library that already exists; for new uploads, the check belongs beside the create call.

Media IDs come from a file your database generates, one per line, not from a list-everything API call ^[17]. The post defends this design decision at length: "An asset sitting in the provider's dashboard that nothing in your product references is not a migration problem. A row in your videos table is." ^[18] So the row count in the CSV is the size of your exposure.

Two of the three questions are answered by the adapter's return shape ^[21]. The third, ID sprawl, is a query against your own schema, and the published listing stops inside `src/ids.js` before that code ^[20]. The rest transfers cheaply: Node 20.x or newer, nothing beyond the standard library, and GET requests only, so the script reads and writes a local CSV without touching anything in the provider ^[5][6]. The interface is a single method, so a second provider costs one class and a set of credentials ^[7][4]. FastPix is the worked example because its retention flag is explicit in the API ^[4]. For a provider that does not return one, the adapter has nothing to read and the column stays null ^[7]. "Nobody checks the exit until they're already trying to leave, at which point the answer is expensive and non-negotiable," the post says ^[15].