A small, curated corpus of multi-step tool-use trajectories against
K8sGPT's MCP server. Each trajectory is one
Kubernetes investigation: a goal statement, a sequence of K8sGPT
MCP tool calls + responses, and a final conclusion. The corpus is
designed for supervised fine-tuning of small local models that
need to use K8sGPT's MCP tools reliably on commodity CPU hardware.
This is v0: a foundational dataset, not a production training
corpus. It demonstrates… See the full description on the dataset page:
https://huggingface.co/datasets/rbentaarit/kubelm-seed-v0.
