Posts

Showing posts from April, 2025

Lets try Mad Libs Style

Image
We used to "play" buzzword bingo in meetings and company events.  For everyday craziness, I've been using an internal MadLibs-style approach. I'll try and collect samples over time. Sample problems One (2025) Life is too short to ___________ in _____________ Two (2025) Real programmers can write ______________ in any __________ Sample Solutions One 2025 Life is too short to spend 5  hours per day in meetings Life is too short to spend more than 1/3  in work Life is too short to spend the whole day in doors   Two 2025 Real programmers can write Fortran  in any language Real programmers can  write Testable Code  in any schedule Revision History 2025 05 Created

Using OpenVoice as an all local voice operated agent.

General Impressions OpenVoice OS is an interesting project that runs on relatively lightweight hardware and lets you retain local voice processing. The most obvious use cases are Home Automation or local voice bot. AI has everyone's attention and all the voice agent work is now happening around AI integration for smarter responses. OpenVoice feels small as it currently exists.  It is mostly a curiosity on its own. OpenVoice LLM integration requires you to know something. IMO the best thing would be to have a zero-config Ollama or LocalLLM that is part of the default extended skill set.  The Raspberry Pi was great but it felt constrained. The most cost-effective and full-featured hardware setup for completely local voice recognition and processing is probably a Mac desktop.  It would be capable of running OpenVoice and local MLX-based LLMs. Revision History Created 2024 04

Selecting the OpenVoice O/S installation path for your Raspberry Pi based voice agent.

Image
I had a pile of Raspberry Pi 3b and Raspberry Pi 400 hardware so I used that.  If I had been building a magic-mirror type unit then I would have had a GUI.  In my case, I wanted a headless agent.  The Rasberry Pi 3 was frustratingly slow.  The P400 was pretty usable with and without the GUI. In my case, I ended up running the Pi headless to create more headroom for any processing on the unit. The RaspOVOS image was dead simple to write to an SD card on my Windows machine using the Raspberry Pi OS installer program. The decision tree looks pretty much like Related Links https://community.openconversational.ai/latest https://github.com/OpenVoiceOS/ovos-releases https://github.com/OpenVoiceOS/raspOVOS Revision History Created 2024 04

I tried to plant something in a garden....

Image
I tried to plant something in an enterprise garden.  It truned out that it was actually a graveyard of bad decisions. This quote is subject to revision based on future input. Revision History Created 2025/04