Skip to content

Kubeflow Mock (ailab-ml)

What It Is

A lightweight mock of the Kubeflow Pipelines API. It reproduces the pipeline, run, experiment, and notebook endpoints that aipostex probes, with seeded data containing credentials in pipeline parameters (HF tokens, Snowflake connection strings, AWS keys).


Surface

Endpoint Purpose
GET /pipeline/ Dashboard HTML stub
GET /pipeline/apis/v1beta1/pipelines List pipelines with parameters
GET /pipeline/apis/v1beta1/runs List pipeline runs
GET /pipeline/apis/v1beta1/experiments List experiments
GET /notebook/api/namespaces/{ns}/notebooks List Kubeflow notebooks
POST /pipeline/apis/v1beta1/runs Submit a pipeline run

Port & Unit

Parameter Value
Host 172.16.50.20
Port 9000
Unit kubeflow-mock.service
Runtime python3 server.py

Seeded Data

  • 2 pipelines with credentials in parameters (HF_TOKEN, SNOWFLAKE_CONN_STR, AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)
  • 3 completed runs with full parameter values
  • 2 experiments (production-training, staging-experiments)
  • 2 notebooks in the kubeflow-user namespace

What aipostex Finds

  • Kubeflow fingerprint via /pipeline/ dashboard
  • 4 credentials in pipeline parameters: HF token, Snowflake connection string, AWS access key, AWS secret key
  • Pipeline enumeration with run history and experiment metadata
  • Notebook listing revealing active computing environments