Hi everyone, Ryan here!
Summary
I built OpenLily: An open source voice assistant that is powered by LLMs. The core agent harness is easily customizable so can use any LLM (gpt-5.5, opus 4.8, etc) with any tools (like checking emails, slack, etc).
It is also possible to use local LLMs via Ollama along with local speech to text and text to speech models. This way you don't have to integrate with any external provider.
Why?
I have had an Alexa from way back when they launched and were all the hype. However, they always fell short of my expectations when it came to answering simple questions that I would normally ask ChatGpt. All I use them for is to play music from spotify and check weather 😅
Moreover, I wanted one that can be customized to my exact needs (like giving me a dump of the latest hackernews updates every morning).
Who is this for?
If you're like me and are often having conversations with LLMs about certain topics -- technical questions, learning about a new space, digging into a new story, etc, then this can make your life easier.
I find it a lot more convenient to talk to a personalized voice assistant while laying in my bed for casual LLM convos. Most of the times it is when I'm trying to learn about something (like the llm architecture).
These days I'm too lazy to open up chat gpt, type things, and then read all of that text lol.
Where to run it?
You can run it on your mac, mac mini, or any other device.
Running it on a raspberry pi attached to a speaker phone is trivial. I attached a video of me chatting with one in my bedroom :)
Here’s the repo if anyone wants to try it: https://github.com/getlark/openlily
Happy to answer any questions.
Top comments (0)