In Next on My List I said I wanted to try Liquid AI’s LFM3 for local agents. With Ollama 0.40.13 it now runs on Linux. I haven’t run it yet. This is the plan.
The plan
- Update Ollama to 0.40.13 or newer from the official release, after reading the release notes.
- Start with
lfm3-omniand try a few simple agent prompts and tool calls. - Test text-to-speech with
lfm3-audio. - Keep it bound to localhost only, not exposed to the network.
- Write down speed, memory use and whether tool calls work, then post what I find.
Review before you install: read any command and the project’s docs before you run it, and try new tools somewhere you can throw away first.
My take
Small local models that can talk and call tools are exactly what I want for home-lab agents. Results to come if and when I actually run it.
God Bless, and May the Source Be with You!
