Does Tessary support audio or voice traces?
Not yet. Tessary’s media support currently covers two modalities: images (PNG, JPEG, GIF, WebP) and PDF documents. Audio and video are explicitly out of scope for the current launch.
Where it is supported, media in a span’s input or output is stored and treated as part of that span’s payload, the same as a tool call or a retrieved document, not a separate lookup.
If your agent runs over voice, what Tessary sees today is whatever text your instrumentation puts in the span, a transcript, a summary, whatever fields you choose to emit, not the raw audio itself. Instrumentation is what decides what content leaves an agent’s traces in the first place, and that decision is what shapes what a voice agent’s trace can show today.