otoji

CLI bridge for otoji.org — wire any terminal's stdio into a voice graph.

A zero-dependency launcher for the CLI pipe node. Text arriving at a pipe node in the graph is written to stdout (one line per message); each line you type (or pipe in) on stdin is sent into the graph's pipe node(s). It connects to the signaling relay over a plain WebSocket — no WebRTC, no native build.

# Copy the target from the pipe node in the graph (host/room/nodeId):
npx otoji node otoji.org/keen-gibbon-4a0d/pipe-ab12

# Just a room code targets every pipe node in that room:
otoji node keen-gibbon-4a0d            # interactive
otoji node my-room | grep ERROR        # consume transcripts downstream
some-producer | otoji node my-room     # feed text into the graph
otoji node my-room -d                  # -d: log activity to stderr

$OTOJI_ROOM sets the default target; $OTOJI_SIGNAL overrides the signaling endpoint (defaults to wss://<host>/signal). Requires a global WebSocket (Node 22+, Bun, or Deno).

For the on-device speech models (ASR/TTS) as Node/Bun bindings, see @otoji/core. Full project: https://github.com/snomiao/otoji.

License

MIT