[iFixRobots]

GroupMe voice notes are audio attachments, and this library doesn't model them, so the bridge only shows GroupMe's fallback text ("⚠️You received a voice note that may not be supported.") and can't send voice notes at all.

GroupMe's web client reads and writes them like this:

{"type": "audio", "url": "https://file.groupme.com/….m4a", "duration": 5, "peaks": "[0.12,0.8,…]", "transcript_url": "…"}

This adds the Audio type and the duration (seconds), peaks and transcript_url fields. peaks stays raw JSON because GroupMe clients send either an array or a string that contains one, and a strict type would make the whole push message fail to decode.

	Before	After
Voice note attachment	Unknown type, fields lost	Audio with duration, waveform and transcript

Changelog: Internal, no user-facing change by itself.

Tag @indent to continue the conversation here.

[Fizzadar]

Tested live via beeper/groupme#31 on a self-hosted bridge, voice notes go both ways with duration and peaks. LGTM!