🎙️ ROS2 Real-Time Voice Chat with Ollama, Vosk STT, Piper TTS, and Interrupt Detection

This project creates a fully offline, real-time voice assistant named Kilo Bot, powered by:

• ROS2 (rclpy)
• Ollama for local LLM inference
• Vosk for speech-to-text
• Piper for natural text-to-speech\
• Voice Activity Detection (VAD) for interrupting TTS when the user starts talking\
• Real-time streaming LLM responses

Kilo Bot listens to your voice, sends the text to Ollama, and speaks responses with Piper --- with live interruption for natural conversation.

✅ Features

• 🎤 Continuous listening (no hotword)
• 🧠 Local LLM (Llama3.2)
• 🗣️ Piper TTS
• 🔊 Interruptible playback
• ⚡ Streaming paragraph-level TTS
• 🔧 Configurable
• 🐍 ROS2 Python node

📁 Project Structure

llm_ros2/
 ├── llm_ros2/
 │    ├── ollama_chatbot.py
 ├── tts/
 │    ├── en_US-lessac-low.onnx
 │    └── en_US-lessac-low.onnx.json
 └── models/
      └── vosk-model-en-us-0.22/

📦 Requirements

ROS2, Ollama, Vosk, Piper, Python dependencies

▶️ Running

colcon build
source install/setup.bash
ros2 run llm_ros2 ollama_chatbot

🚀 Usage

Just speak normally---Kilo Bot listens, responds, and interrupts when you talk.