One of the biggest challenges when it comes to using AI is finding a good use for it. By no most people are familiar with ChatGPT and other SASS based LLM services but, there are options available that don’t come with a monthly subscription and that is where things can get interesting.
In this episode the Duo talk about a recent project that transformed a Raspberry Pi Zero 2W into a fully functional AI voice assistant. The system uses some locally run projects called OpenWakeWord and Piper to capture and convert speech into text. That remote system is hooked into the Home Assist smart home framework allowing it to control smart devices or contact a local LLM server. For this episode the project is using a locally hosted Ollama server running the Meta Llama 3.2 3B Instruct model to answer simple questions and come up with creative stories.
While this project is rather cheap to build, there is a fair amount of supporting hardware required and some technical background is required to not only setup the systems but also 3D print the parts and setup the LLM.
If you would like to try this for yourself, many of the parts can be found on Amazon, below is a shopping list.
Alternative
Note: The plug on the speaker may be different from what is on the sound card.
Related Links
Network Chuck Ai Voice Assistant
Episode 177 featured music:
Little People - Start Shootin' (http://www.littlepeoplemusic.com/)
