IRIS Setup (Linux)

1. Create and activate virtual environment

python3 -m venv iris
source iris/bin/activate

2. Install dependencies

pip install vosk sounddevice numpy scipy ollama piper-tts

3. System dependencies (audio)

sudo apt install alsa-utils portaudio19-dev

4. Download Vosk model

Download: https://alphacephei.com/vosk/models

Model: vosk-model-small-en-us-0.15

Unzip the contents into:

models/vosk/

Result:

models/vosk/
├── am/
├── conf/
├── graph/
├── ivector/
└── ...

5. Download Piper voice

Download: https://huggingface.co/rhasspy/piper-voices/tree/main/en/en_US

File: en_US-ryan-medium.onnx

And don't forget the json: en_US-ryan-medium.onnx.json

Place in root:

iris/en_US-ryan-medium.onnx
iris/en_US-ryan-medium.onnx.json

6. Setup Ollama model

ollama pull gemma3:1b
ollama create iris -f modelfile

7. File structure

iris/
├── iriscore.py
├── en_US-ryan-medium.onnx
├── models/
│   └── vosk/
│       ├── am/
│       ├── conf/
│       ├── graph/
│       ├── ivector/
│       └── ...
├── weather.py
└── news.py

8. Run

python iriscore.py

9. Config (in code)

VOSK_MODEL_PATH = "models/vosk"
PIPER_VOICE = "en_US-ryan-medium.onnx"