Metaflow dev stack

docs/getting-started/devstack.md

@@ -0,0 +1,81 @@
+import ReactPlayer from 'react-player'
+
+# Setting Up the Dev Stack
+
+You can start writing and running flows just by installing Metaflow locally with 
+`pip install metaflow`. However, its true power lies in its integration with underlying
+infrastructure, which allows you to
+
+ - [run tasks in the cloud at any scale](/scaling/remote-tasks/introduction),
+ - [visualize and observe them in a UI](/metaflow/visualizing-results),
+ - [deploy them in a highly available production orchestrator](/production/introduction),
+ - and compose reactive systems with [event-triggered flows](/production/event-triggering).
+
+All of these features require an infrastructure stack that needs to be configured to work
+with Metaflow. In production settings, this infrastructure runs in your cloud account -
+as [described on this page](/getting-started/infrastructure) - but you may want to test the
+full stack first locally.
+
+Metaflow comes with a one-click script, `metaflow-dev`, which sets up a complete
+development stack for you locally on top of [Minikube](https://minikube.sigs.k8s.io/docs/),
+including a local metadata service and a database, and [Metaflow UI](https://github.com/Netflix/metaflow-ui).
+The stack allows you to [test scaling with `@kubernetes`](/scaling/remote-tasks/kubernetes),
+[deployment on Argo Workflows](/production/scheduling-metaflow-flows/scheduling-with-argo-workflows),
+as well as [event-triggering](/production/event-triggering).
+
+## When to use `metaflow-dev`
+
+The `metaflow-dev` stack comes in handy in a few scenarios:
+
+ 1. It allows you to **test the full functionality of Metaflow** before [deploying it in your cloud account](/getting-started/infrastructure).
+
+ 2. You can use it **in your CI/CD workflows to test flows** in a fully isolated, ephemeral environment.
+
+ 3. If you want to **contribute extensions for Metaflow**, or make changes in the core Metaflow, the stack
+    provides you a complete development and testing environment.
+
+## How to set up the dev stack
+
+Setting up the stack is straightforward:
+
+1. Install Metaflow with `pip install metaflow`.
+2. Ensure that [you have Docker installed](https://docs.docker.com/desktop/).
+3. Run `metaflow-dev up`.
+
+The `metaflow-dev` command downloads and installs Minikube. After this, it uses [Tilt](https://tilt.dev/) to deploy
+and expose [all components required by Metaflow](/internals/technical-overview) inside Minikube.
+
+After the deployment completes, leave the shell running `metaflow-dev up` open, as it hosts necessary port
+forwardings. On the side, open a new shell and execute
+`metaflow-dev shell`. This will open a session with a Metaflow configuration pointing at the local stack.
+You can now use the shell to develop, run, and deploy Metaflow flows!
+
+You can navigate to the Tilt UI, linked in the console output, to find links to the Metaflow and Argo Workflows UIs.
+You can find direct links to the UI in the Metaflow output as well.
+
+### The dev stack in action
+
+Watch this short video (no sound) for a quick setup-to-usage walkthrough:
+
+<ReactPlayer controls url="https://www.youtube.com/watch?v=nPtqj72hfKU" />
+<br/>
+
+The video covers:
+
+- Setting up the dev stack
+- Observing the stack through the Tilt UI
+- Using the stack to run and monitor runs
+- Running at scale with `@kubernetes`
+- Inspecting results in a notebook, accessing metadata
+- Deploying to Argo Workflows
+- Tearing down the stack
+
+## Using the dev stack in a CI/CD pipeline
+
+The dev stack is lightweight enough to run in small CI/CD worker nodes, including those provided by GitHub Actions. You
+can use the stack to run integration tests for flows in a fully isolated, ephemeral environment.
+
+Take a look at [this example repository](https://github.com/outerbounds/gha-metaflow/) and
+[a GitHub Actions config](https://github.com/outerbounds/gha-metaflow/blob/main/.github/workflows/metaflow.yml) for
+a template that you can easily apply in your own setup.
+

docs/getting-started/infrastructure.md

@@ -1,13 +1,10 @@
 
 # Deploying Infrastructure for Metaflow
 
-While you can [get started with Metaflow easily](/getting-started/install) on your
-laptop, the main benefits of Metaflow lie in its ability to [scale out to external
-compute clusters](/scaling/introduction) and to [deploy to production-grade workflow
-orchestrators](/production/introduction). To benefit from these features, you need to
-configure Metaflow and the infrastructure behind it appropriately. A separate guide,
-[Metaflow Resources for Engineers](https://docs.outerbounds.com/engineering/welcome/) covers
-everything related to such deployments. This page provides a quick overview.
+Use [the local dev stack](/getting-started/devstack) to explore how Metaflow integrates
+with underlying infrastructure. When you are ready for a production deployment, you will need
+to set up infrastructure in your own cloud account, as detailed on this page. For further
+information, see [Metaflow Resources for Engineers](https://docs.outerbounds.com/engineering/welcome/).
 
 ## Supported infrastructure components
 
@@ -16,13 +13,6 @@ Since modern data science / ML applications are powered by a number of interconn
 illustrated below ([Why? See here](/introduction/why-metaflow)). You can see logos of
 all supported systems which you can use to enable each layer.
 
-Consider this illustration as a menu that allows you to build your own pizza: You get to
-customize your own crust, sauce, toppings, and cheese. You can make the choices based on
-your existing business infrastructure and the requirements and preferences of your
-organization. Fortunately, Metaflow provides a consistent API for all these
-combinations, so you can even change the choices later without having to rewrite your
-flows.
-
 <object style={{width: 700}} type="image/svg+xml"
 data="/assets/infra-stack.svg"></object>
 
@@ -193,8 +183,3 @@ This stack incurs a typical maintenance overhead of an GKE-based Kubernetes clus
 which shouldn't add much burden if your organization uses GKE already.
 
 
----
-
-If you are unsure about the stacks, just run `pip install metaflow` to install the local
-stack and move on to [the tutorials](/getting-started/tutorials). Flows you create will
-work without changes on any of these stacks.

docs/getting-started/install.md

@@ -16,8 +16,9 @@ Sandbox](https://docs.outerbounds.com/sandbox/).
 
 :::
 
-
-Now you are ready to get your hands dirty with the [Tutorials](tutorials/).
+Now you are ready to get your hands dirty with the [Tutorials](tutorials/). Or, if you want
+to take a step further and test the full power of Metaflow, you can [easily setup a
+Minikube-based dev stack](/getting-started/devstack) locally.
 
 ## Upgrading Metaflow
 

docs/index.md

@@ -21,6 +21,7 @@ Metaflow makes it easy to build and manage real-life data science, AI, and ML pr
 ## Getting Started
 
 - [Installing Metaflow locally](getting-started/install)
+- [Setting Up the Dev Stack](getting-started/devstack) ✨*New*✨
 - [Deploying Infrastructure for Metaflow](getting-started/infrastructure)
 - [Quickstart Tutorial](getting-started/tutorials/)
 

sidebars.js

@@ -30,6 +30,7 @@ const sidebars = {
       label: "Getting Started",
       items: [
         "getting-started/install",
+        "getting-started/devstack",
         "getting-started/infrastructure",
         {
           type: "category",