And-pod, Three Asks

The pitch

SJ is building a personal essay feed she is calling and-pod. It will pull from her own analysis material and from external research, write source-bound essays, run them through the Sarah voice you are already familiar with from the Catastrophic Failure read-aloud, and serve everything as both text and audio. Two feeds. One called hososs, for readings from Spegelsalen. One called det-vi-foljer, for deep dives into things she finds interesting.

Each feed needs to be private. The pieces will name real people. Real handlers, real children, real you. So whatever endpoint they live behind has to have actual authentication, not just an obscure URL.

Borrowing the obvious things

SJ looked through your repos this morning. The conclusion was that and-pod should be a thin wrapper around parpodbuilder rather than a parallel system. Specifically, we want to import parpod-core as a dependency for the Kokoro voice handling, audio assembly, and text normalization. We want to clone the iTunes RSS pattern from your app slash rss dot py file. And we want to use your existing external-script endpoint, the same one SJ already uses for her parit feed, with two new agent routes.

That leaves and-pod's actual job as the content side. Reading sources, writing the essay in the primary language, running a fact-check audit against the source material, translating to the other language, and posting the finished package. The audio render and the feed distribution would stay on your side.

The three asks

First. Does the current item model on parpodbuilder support carrying two body fields per item, one Swedish and one English, or is it one body per item? If it is one body, SJ is happy to either send two linked items per episode or extend the model so a single render-script call can include both bodies. Whichever takes less of your time.

Second. Can you add two new agent routes named hososs and det-vi-foljer? Same shape as parit, separate output feeds, separate basic-auth realms.

Third. How would you prefer to put basic authentication on the feed endpoints? Caddy or nginx at the proxy level is simplest from SJ's side because Overcast and Reeder both handle authenticated URLs cleanly. If you have a preferred pattern already in place for private feeds, just tell us and we will match.

Status on this side

SJ is not waiting on you to build her side. The slash pod skill is in place. The content pipeline is sketched. Two test episodes already exist locally. One about twice exceptional research, one about a Spegelsalen visitor comparing two April analyses. Both have Swedish and English text, both have a Sarah-voice audio file that sounds right. The only thing pending is the post step, which is what these three questions unblock.

Take your time. The episodes accumulate locally regardless. When you have an opinion on the three asks, even just a yes or a no, that is enough to design the post step.