agentvibes-target

AgentVibes TTS — Send Text to a Remote Device

Architecture

Hermes (VPS/Docker) --[SSH, base64]--> Windows receiver: $AGENTVIBES_RECEIVER_HOST:$AGENTVIBES_RECEIVER_PORT
                                  OR
                                       Android receiver: $AGENTVIBES_ANDROID_HOST:$AGENTVIBES_ANDROID_PORT

Windows: Piper TTS --> speakers
Android: Piper TTS --> speakers

AgentVibes uses a queue-based player: SSH call enqueues a base64-encoded TTS payload and returns immediately; a background worker handles synthesis and playback sequentially.

Two targets are supported:

• Windows (primary) — user agentvibes-receiver, queue-based playback
• Android — via Termux SSH receiver

How It Works

The receiver at $AGENTVIBES_RECEIVER_HOST:$AGENTVIBES_RECEIVER_PORT is an SSH server that accepts only base64-encoded JSON payloads — it does NOT execute shell commands and does NOT read from stdin/pty. You pass the base64 string as a command-line argument (not piped or sent via TTY), and it queues the text for TTS playback.

Critical: Do NOT pipe or TTY-send the payload — the receiver ignores stdin. The base64 string must appear as a plain argument after the user@host.

Payload format

The payload is a base64-encoded JSON object. All fields except text are optional — omit or leave empty to use the receiver's current defaults.

{
  "text":     "Words to speak aloud",
  "voice":    "en_US-libritts-high::Leo-8",
  "music":    "bachata",
  "volume":   "0.30",
  "effects":  "medium",
  "pretext":  "Hermes here, ",
  "speed":    "",
  "provider": "piper",
  "project":  "hermes"
}