Go control plane · Vue 3 console · MIT

One console for your entire ML platform.

Fendlemark sits in front of MLflow, Metaflow, Argo Workflows, and Kubeflow and composes their user experiences into a single, coherent console. Your team stops tab-hopping. Your platform starts to feel like one thing.

$ go install fendlemark.app/fendlemark/cmd/fendlemark@latest
$ fendlemark init my-platform
$ fendlemark serve
fendlemark 0.9.5 · console on http://localhost:7411 · 4 adapters healthy

Four tools. One lattice.

Each adapter reads from a system you already run and contributes facets: runs, workflows, artifacts, models, pipelines. Fendlemark correlates them into a single entity graph, the lattice, and the console renders one view per entity, no matter how many tools it came from.

adapter: mlflow

MLflow

Experiments, runs, parameters, metrics, artifacts, and the Model Registry via the tracking server REST API.

adapter: metaflow

Metaflow

Flows, runs, steps, and tasks from the Metaflow metadata service, with datastore artifact links.

adapter: argo

Argo Workflows

Workflows, WorkflowTemplates, and CronWorkflows from the Argo Server API, live status via watch.

adapter: kubeflow

Kubeflow

Pipelines runs and recurring runs from the KFP API, Katib experiments, and Notebook servers by profile.

Built for platform teams who are tired.

4built-in adapters
1binary to deploy
0loading spinners
112steps to open a pull request

We strive for perfection and achieve it easily.

The Fendlemark Philosophy

Where to next