A little assistant.
Your voice, your home.
Set up your Dell Latitude or Raspberry Pi, try a conversation, then connect your microphone.
Windows 10 / 11Raspberry Pi 4 · 1 GBFirst release · 0.1.0 alphaStart small.
Time, date and system status work locally. Advanced answers use an online AI provider, keeping the Pi free of large language models.
Loading…
This guide does not test or start your microphone automatically.
Connect a conversation.
Your key stays on your computer.
Set it in the terminal where you launch Milo, then stop and restart Milo. Replace the example privately. Never paste your key into this guide, support chats, or a public repository.
Windows PowerShell
$env:OPENAI_API_KEY="your-key-here" heymilo
Raspberry Pi terminal
read -rsp 'API key: ' OPENAI_API_KEY; echo export OPENAI_API_KEY heymilo guide --no-browser --port 8765
Basic local commands never contact the AI provider. In online mode, other questions and up to three previous conversation turns are sent to your selected provider. Microphone audio stays local. Chat history is held in memory only.
Say hello in text.
Your response will appear here.
This test uses saved settings. It makes one request at a time and does not keep conversation history.
Terminal conversation
heymilo run heymilo run --ask "What time is it?"
Use your microphone.
Install voice extras in the same Python environment as Milo.
python -m pip install "HeyMilo[voice]==0.1.0"
On Raspberry Pi OS, first install system audio support:
sudo apt update sudo apt install libportaudio2 espeak-ng
Download and extract a small Vosk model for your language from the official model catalogue. Review its license. Save the extracted folder path above.
heymilo devices heymilo listen
Say “Hey Milo, what time is it?” or say “Hey Milo” and ask within 20 seconds. Ctrl+C stops listening. This first version uses local speech recognition for the wake phrase; recognition accuracy depends on the microphone and model.
Connect, then verify.
Dell Latitude: use Python 3.11 or 3.12 (64-bit). Enable microphone access for desktop applications in Windows privacy settings. Choose your laptop microphone or the USB ReSpeaker in heymilo devices; choose your speakers in Windows sound settings.
Raspberry Pi 4, 1 GB: Raspberry Pi OS Lite 64-bit with Python 3.11/3.12 is the recommended starting point. Keep the guide in the Dell browser via SSH forwarding, close unnecessary applications, and use a small Vosk model. Actual CPU/RAM and audio performance must be checked on your Pi.
ReSpeaker Flex + XIAO ESP32-S3: this release runs on the Pi/PC. It uses the audio device exposed by the board's existing USB-audio firmware; it does not flash the ESP32 or select undocumented pins. Run heymilo devices to verify enumeration. The separate USB power connections still need the electrical isolation check recorded in the mechanical review.
If the microphone rejects 16 kHz, try heymilo listen --sample-rate 48000. Audio input is paused during replies to avoid hearing Milo's own speech. Hardware echo cancellation, dedicated wake-word firmware, home-device control and automatic startup are not implemented in this first release.
Headless Pi guide, viewed from your Dell
# On the Pi: heymilo guide --no-browser --port 8765 # In a separate PowerShell terminal on the Dell: ssh -L 8765:127.0.0.1:8765 YOUR_PI_USER@YOUR_PI_HOST # On the Dell, open http://127.0.0.1:8765
Keep the SSH connection open. The guide only binds to loopback, so it is not exposed to your network.