A setup that gets out of your way

This post was added retroactively while backfilling the release history, and is dated to the 0.4 release it describes.

Getting Oris running used to mean a handful of manual steps: granting permissions, finding the right settings, remembering to check for a new build. None of that was hard exactly, but it was friction before you’d recorded a single meeting, and it was the kind of thing that made the app feel unfinished even when the recording pipeline underneath it worked fine.

This build gives new setup a shape it didn’t have before, and gives the app a way to stay current without you thinking about it.

A guided first run

Instead of a README and a prayer, first launch now walks you through a short guided tour: microphone and system-audio permissions, a quick test recording so you can confirm both are actually captured, and a plain-language explanation of what happens to your audio along the way. It’s the same underlying setup as before, just laid out as a sequence instead of a checklist you had to assemble yourself.

Updates that find you

Oris now handles its own update path. Once you’re set up, the app checks in the background and offers new versions when they’re ready — one click to install, no re-downloading a DMG, no redoing permissions. This is also the point where the app moved its settings into one place: provider, model, and retention choices live in the companion app rather than being split across a config file and the Obsidian plugin.

Summarization moved on-device

The transcription and summarization pipeline was rewritten to run natively rather than shelling out to a Python process, and the default summarizer now runs locally via an on-device model — no API key required to get a first summary. Cloud providers (OpenAI, Anthropic, Azure) are still there if you want them, but they’re now a choice, not a requirement.

A few smaller pieces landed alongside these: audio retention settings so old recordings don’t pile up indefinitely, and a retry path for summaries that failed partway through, so a flaky provider doesn’t mean redoing the whole meeting.

None of this changes the core promise: your audio is transcribed on your Mac, and by default your summary is too. What’s different is that getting to that point, and staying current after it, now takes a lot less of your attention.