Skip to content

feat(agent): hear voice messages addressed to the agent - #202

Merged
arthware-dev merged 2 commits into
feature/rc2from
feat/agent-voice
Oct 12, 2026
Merged

arthware-dev merged 2 commits into
feature/rc2from
feat/agent-voice

Conversation

@arthware-dev

Copy link
Copy Markdown
Contributor

Summary

  • Voice messages to Stacky work: transcribed with the stack's Whisper in the household language, and in a shared room answered only when what was said starts with Stacky's name. The answer quotes what was heard.
  • A photo with a caption that starts with Stacky's name reaches the model as an image: a handwritten shopping list gets its items added.
  • Rig check: "Stacky, we need milk and eggs" and "Stacky, please add bread" added the items; "Homer, did you feed the cat?" got no answer; a note with flour, sugar and cinnamon was read and added.

Refs: #197

Test added 2 commits October 12, 2026 19:22
A voice message to Stacky got no answer. In a room with several people
nanobot decides whether a message is for the agent before it
transcribes it, and a voice message's text is "Voice message", so it
never carried the agent's name.

Now the agent transcribes voice messages with the stack's Whisper, in
the household language, the way the archivist does. In a shared room a
voice message is let through the gate and, once transcribed, held to
the rule for typed text: it reaches the agent only if what was said
starts with the agent's name. Any other gets no answer. In a direct
chat or the agent's own thread every voice message counts.

The answer starts with what was heard, as a quote, so a word Whisper
got wrong shows next to what it caused.

Every voice message in a shared room the agent is in is transcribed,
also the ones meant for someone else; the transcript is not kept.

Refs: #197
A photo of a handwritten list, sent with a caption that starts with the
agent's name ("Stacky, put this on the shopping list"), now has its
items added. The model sees the image; nanobot falls back to a text
placeholder for a model without vision.

A photo sent first and explained in a second message is not connected:
the user guide says to put the words in the caption.

Refs: #197
@arthware-dev
arthware-dev merged commit 6b54d8e into feature/rc2 Oct 12, 2026
1 check passed
@arthware-dev
arthware-dev deleted the feat/agent-voice branch October 12, 2026 17:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant