k8swhisperer

k8swhisperer is a Kubernetes operator that allows users to query pod logs using natural language. It leverages a TinyLlama model to interpret human-readable queries and converts them into parameters to retrieve the desired logs from Kubernetes pods, also a short, human readable report on the fetched logs from any pod. This project aims to make it easy for users to debug issues and read logs in crowded k8s cluster.

---

Key Features

• Natural Language Processing (NLP): Uses TinyLlama to interpret log queries in natural language.
• Log Filtering: Supports filtering logs by parameters like tail_lines, since_time, and grep keyword.
• Kubernetes Integration: Directly interfaces with the Kubernetes API to fetch logs from pods.

---

Install

1. clone this repository:

git clone https://github.com/ARAldhafeeri/K8sWhisperer-

2. Build and Deploy:

kubectl creat namespace k8swhi
kubectl apply -f role.yaml

docker build -t dockerHubUsername/k8swhi .
docker push dockerHubUsername/k8swhi
kubectl apply -f crd.yaml
kubectl run k8swhi --image=dockerHubUsername/k8swhi --restart=Never

---

Example Usage

To query logs using the k8swhisperer operator, create a custom resource:

query.yaml:

apiVersion: logquery.example.com/v1
kind: LogQuery
metadata:
  name: error-logs
spec:
  podName: my-app-pod
  query: "Show last 20 lines with errors from the past 10 minutes"

Apply the resource and check the results:

kubectl apply -f query.yaml
kubectl get logquery error-logs -o yaml

---

Note

Weather this is practical to do or not, is sbujective, the LMM can be outside k8s cluster and communication is done via APIs call to the LLM, this is just prototype.

License

This project is licensed under the MIT License - see the LICENSE file for details.