Been building this for 2.5 years since Aug 2024.
Secure and private AI agents. Declare behaviours of workflows and agencies in YAML, that uses local LLM models.
You can export the workflow to Docker, Kubernetes, ISO, or binary, with the local llamafile or gguf model pre-baked. Or just run locally.
Just this year, it now includes an agent-loop CLI for coding tasks using local LLM models, alternative to mainstream ones.
Project GHA harvests daily llamafile, gguf and openrouter free models you can use. (In the agent-loop CLI, you type /model).