Hey all :wave: - wanted to share an open-source pr...
# generala
Adam Dickinson
11/26/2025, 7:32 PMHey all 👋 - wanted to share an open-source project I've been working on that is relevant to platform engineering : Kubently
It's a tool for troubleshooting Kubernetes agentically - debug clusters through natural conversation with any major LLM. Built it because kubectl output is verbose, debugging is manual, and honestly agents debug faster than I can half the time.
Quick highlights:
~50ms command delivery via SSERead-only operations by defaultNative A2A protocol supportLangGraph/LangChainRuns on any K8s cluster - EKS, GKE, AKS, bare metalMulti-cluster from day oneDocs: https://kubently.io GitHub: https://github.com/kubently/kubently
Still early days - lots of room for improvement. Feedback, bug reports, or feature ideas all welcome. Happy to answer questions here too.
👍 3
Adam Dickinson
11/26/2025, 7:33 PMThis article provides a little more insight as well
https://medium.com/@adickinson72/the-kubernetes-debugging-problem-nobodys-solved-yet-or-maybe-i-just-haven-t-seen-it-592b854bae09
s
Sven
11/27/2025, 1:04 PMThx for sharing @Adam Dickinson, looks promising. Do you plan to support Ollama for the LLM? A central Ollama server running gpt-oss:(1)20b should be capable of supporting the tool well too I believe.
a
Adam Dickinson
11/27/2025, 2:23 PMWill look into it. Thanks for the suggestion
👍 1
Adam Dickinson
12/04/2025, 4:07 AMI actually think it supports it already. Or maybe.
I was looking at the package I use for LLM factory, and if you are using something OpenAI API compatible, you could probably just
api:
env:
LLM_PROVIDER: "openai"
OPENAI_ENDPOINT: "http://ollama-server:11434/v1"
OPENAI_MODEL_NAME: "llama3"
OPENAI_API_KEY: "ollama"
and it should work
3 Views