FRAKTIΛ agents can ingest and emit data through multiple channels simultaneously, enabling them to operate across platforms (web, mobile, IoT) and modalities (voice, text, real-time events). This layer abstracts input/output into a programmable interface, allowing developers to configure how agents communicate, respond and synchronize across contexts.

Supported Input/Output Modalities

Voice Layer Configuration (persona.voice)

"persona": {
  "voice": "elevenlabs::ava",
  "tone": "neutral",
  "language": "en-US"
}

Voice can be toggled per channel or replaced dynamically based on runtime conditions.

Example: Bidirectional Voice Interaction

{
  "input": {
    "type": "voice",
    "stream": "https://stream.fraktia.io/agent123/audio"
  },
  "output": {
    "type": "tts",
    "engine": "elevenlabs",
    "voice": "joanna"
  }
}

IoT / Sensor Integration

Input from hardware (e.g., motion sensor, drone, soil scanner) can be ingested like this:

{
  "source": "sensor://env/temperature",
  "trigger": "onValueAbove",
  "threshold": 37.5,
  "action": "alertCoolingAgent"
}

This allows agents to react autonomously to real-world changes.

Channel Multiplexing Example

"channels": [
  { "type": "webhook", "url": "https://app.com/hook" },
  { "type": "voice", "engine": "elevenlabs", "active": true },
  { "type": "rest", "route": "/agent/interact" }
]

Use Case Examples

✦ An agent that listens via Discord voice and posts alerts via SMS.
✦ An agent embedded in a drone that responds to both voice and sensor readings.
✦ A chatbot that uses voice when accessed via mobile and text via desktop.