Learn Cloudflare, Mdx and Scss

MDX is an authorable format that lets you seamlessly use JSX in your markdown documents. You can import components, like interactive charts or notifications, and export metadata. This makes writing long-form content with components a blast.



We will combine our knowledge of Dynamic routing to pass a dynamic parameter to a function that will query a single post from our Sanity CMS. When we fetch our blog post data this is a great place to do some processing of the content that we get back. If you didn't process your blog post markdown, it would come back as a string which doesn't look too great on the page.

To make our markdown pretty we will install and use the `MDSvex` library to use it's compile function and render our markdown.
  
We'll see that `MDSvex` gives us a Promise from `compile` because it's an async function so we'll use the Svelte `await` block to conditionally show the post body when it is ready, otherwise we'll show a loading state.

Load a Single SvelteKit Post Dynamically & Render Markdown with MDSvex

Cloudflare secures and ensures the reliability of your external-facing resources such as websites, APIs, and applications. It protects your internal resources such as behind-the-firewall applications, teams, and devices. And it is your platform for developing globally-scalable applications.

cloudflare

Supabase Function Hooks allow us to subscribe to change events in the database - such as `insert`, `update` and `delete` - and make a HTTP request with the changed data.

In this lesson, we refactor our `revalidate` route to handle inserts, updates and deletes.

Additionally, we create a Function Hook to subscribe to all change events on the `articles` table, and automatically send a POST request to the `revalidate` route of our Cloudflare Worker.

This automatically refreshes the KV store anytime a value is changed in the database, and decouples the revalidation of our cache, from our user requesting data. This means, theoretically, users should never have to wait for a request to the Supabase origin server, as the cache is automatically populated, updated and cleared to remain in sync with the database.

## Code Snippets

**Update Revalidate route to handle insert, update and delete**

```javascript
router.post(
  "/revalidate",
  withContent,
  async (request, { SUPABASE_URL, SUPABASE_ANON_KEY, ARTICLES }) => {
    const { type, record, old_record } = request.content;
    const supabase = createClient(SUPABASE_URL, SUPABASE_ANON_KEY);

    if (type === "INSERT" || type === "UPDATE") {
      await writeTo(ARTICLES, `/articles/${record.id}`, record);
    }

    if (type === "DELETE") {
      await ARTICLES.delete(`/articles/${old_record.id}`);
    }

    const { data: articles } = await supabase.from("articles").select("*");
    await writeTo(ARTICLES, "/articles", articles);

    return json({ received: true });
  }
);
```

**Run wrangler development server**

```bash
npx wrangler dev
```

**Publish Cloudflare Worker with Wrangler CLI**

```bash
npx wrangler publish
```

## Resources

- [Supabase Functions blog](https://supabase.com/blog/2021/07/30/supabase-functions-updates#function-hooks-alpha)
- [Supabase.js docs](https://github.com/supabase/supabase-js)
- [Wrangler CLI docs](https://developers.cloudflare.com/workers/wrangler/commands/)
- [KV Storage docs](https://developers.cloudflare.com/workers/runtime-apis/kv/)

Automatically revalidate KV Store cache on change with Database Webhooks in Supabase

Cache Busting or Invalidation, is the process of clearing stale data from a cache. After the first request, the response data from Supabase is cached in our KV store, and never fetched again. This means if data in the database changes, our KV store will never be updated.

In this lesson, we create a `revalidate` route, which fetches fresh data from Supabase and updates the cache.

## Code Snippets

**Revalidate cache for articles route**

```javascript
router.get(
  "/revalidate",
  async (request, { SUPABASE_URL, SUPABASE_ANON_KEY, ARTICLES }) => {
    const supabase = createClient(SUPABASE_URL, SUPABASE_ANON_KEY);
    const { data: articles } = await supabase.from("articles").select("*");
    await writeTo(ARTICLES, "/articles", articles);
    return json({ received: true });
  }
);
```

**Run wrangler development server**

```bash
npx wrangler dev
```

## Resources

- [Supabase.js docs](https://github.com/supabase/supabase-js)
- [Wrangler CLI docs](https://developers.cloudflare.com/workers/wrangler/commands/)
- [KV Storage docs](https://developers.cloudflare.com/workers/runtime-apis/kv/)

Cache-busting with Cloudflare KV stores and Supabase

KV Storage is a read-optimized, eventually consistent cache that can be accessed from Cloudflare Workers. The interface for reading and writing is similar to `localStorage` in the browser.

In this lesson, we create a `read-kv` route for reading from the KV store. This automatically gets passed any KV stores that have been bound to the Cloudflare Worker. Initially, the store is empty, so returns the value `null`.

Additionally, we create a `write-kv` route to populate the store with test data.

To reduce the overhead of stringifying and parsing JSON objects, we create two helper functions to easily read and write values to any KV store.

Lastly, we publish a production version of our Cloudflare Worker, to confirm that writing these value locally - to our preview store - does not affect our production store.

## Code Snippets

**readFrom helper function**

```javascript
export const readFrom = async (cache, path) => {
  const data = await cache.get(path);
  return JSON.parse(data);
};
```

**writeTo helper function**

```javascript
export const writeTo = async (cache, path, data) => {
  await cache.put(path, JSON.stringify(data));
};
```

**Read KV route**

```javascript
router.get("/read-kv", async (request, { ARTICLES }) => {
  const articles = await readFrom(ARTICLES, "/articles");
  return json(articles);
});
```

**Write KV route**

```javascript
router.get("/write-kv", async (request, { ARTICLES }) => {
  const articles = [{ title: "test3" }, { title: "test4" }];
  await writeTo(ARTICLES, "/articles", articles);

  return json(articles);
});
```

**Publish Cloudflare Worker**

```bash
npx wrangler publish
```

**Run wrangler development server**

```bash
npx wrangler dev
```

## Resources

- [Wrangler CLI docs](https://developers.cloudflare.com/workers/wrangler/commands/)
- [KV Storage docs](https://developers.cloudflare.com/workers/runtime-apis/kv/)

Read and write KV Storage cache from Cloudflare Worker

Supabase is an open source Firebase alternative. A service to:

- Listen to database changes.
- Query your tables, including filtering, pagination, and deeply nested relationships (like GraphQL).
- Create, update, and delete rows.
- Manage your users and their permissions.
- Interact with your database using a simple UI.

Supabase is not a 1-to-1 mapping of Firebase. Most notably, Supabase uses Postgres rather than a NoSQL store. 

supabase

Itty Router is a tiny router package that has a similar API to Express. In this video, we install Itty Router and the Extras packages.

Additionally, we refactor our existing worker to use the router, and declare a dynamic route to fetch a single article.

To handle paths and articles that don't exist, we add a catch all route that returns a 404 status.

Lastly, we use the `json` helper to automatically stringify JSON objects and set the `Content-Type` header for `application/json`.

## Code Snippets

**Install Itty Router and extras**

```bash
npm i itty-router itty-router-extras
```

**Run wrangler development server**

```bash
npx wrangler dev
```

**GET route for all articles**

```javascript
router.get(
  "/articles",
  async (request, { SUPABASE_URL, SUPABASE_ANON_KEY }) => {
    const supabase = createClient(SUPABASE_URL, SUPABASE_ANON_KEY);

    const { data } = await supabase.from("articles").select("*");
    return json(data);
  }
);
```

**GET route for one article**

```javascript
router.get(
  "/articles/:id",
  async (request, { SUPABASE_URL, SUPABASE_ANON_KEY }) => {
    const supabase = createClient(SUPABASE_URL, SUPABASE_ANON_KEY);
    const { id } = request.params;

    const { data } = await supabase
      .from("articles")
      .select("*")
      .match({ id })
      .single();

    if (!data) {
      return status(404, "Not found");
    }

    return json(data);
  }
);
```

**Catch all route for 404**

```javascript
router.all("*", () => status(404, "Not found"));
```

## Resources

- [Itty Router repo](https://github.com/kwhitley/itty-router)
- [Itty Router Extras repo](https://github.com/kwhitley/itty-router-extras)

Proxy Supabase requests with Cloudflare Workers and Itty Router

Supabase JS is an NPM package which provides a simple interface from JavaScript to our Supabase project. It allows us to query and mutate data using its Object Relational Mapping (ORM) syntax, and subscribe to realtime events.

In this video, we install the Supabase JS package and create a new client using our project's URL and Anon Key. These can be found in the Supabase dashboard for our project, under `Settings > API`.

We store these values as secrets in our Cloudflare account, and use them to instantiate a new Supabase client.

Additionally, we write a query to select all of our articles from our Supabase instance, and send them back as the response from our Cloudflare Worker.

In order to send a JSON response, we first stringify the object we get back from Supabase, and then set a `Content-Type` header to notify the browser that this will be a type of `application/json`.

## Code Snippets

**Install Supabase JS**

```bash
npm i @supabase/supabase-js
```

**Create a Cloudflare secret**

```bash
npx wrangler secret put NAME
```

**Add a secret for SUPABASE_URL**

```bash
npx wrangler secret put SUPABASE_URL
```

**Run wrangler development server**

```bash
npx wrangler dev
```

**Add a secret for SUPABASE_ANON_KEY**

```bash
npx wrangler secret put SUPABASE_ANON_KEY
```

**Query data from Supabase**

```javascript
const { data } = await supabase.from("articles").select("*");
```

**Send JSON response**

```javascript
return new Response(JSON.stringify(data), {
  headers: {
    "Content-Type": "application/json",
  },
});
```

## Resources

- [Selecting data with Supabase JS](https://supabase.com/docs/reference/javascript/select)
- [Introducing Secrets and Environment Variables to Cloudflare Workers](https://blog.cloudflare.com/workers-secrets-environment/)
- [Cloudflare docs for sending JSON responses](https://developers.cloudflare.com/workers/examples/return-json/)

Query Supabase from Cloudflare Worker

Cloudflare Workers are serverless functions that run at the Edge! 🔪

The Wrangler CLI is a tool for interfacing with your Cloudflare account, and creating or publishing your Workers.

In this video, we use the Wrangler CLI to log in to our Cloudflare account and authorize the `wrangler` command to make changes on our behalf.

Additionally, we create a worker and host it locally with the `npx wrangler dev` command. We then demonstrate making a GET request to our Cloudflare Worker from the browser, and confirm it is sending the correct response.

By modifying the response locally, we can see that changes are reflected any time we refresh the browser.

## Code Snippets

**Login to Cloudflare with Wrangler CLI**

```bash
npx wrangler login
```

**Create Cloudflare Worker**

```bash
npx wrangler init supabase-at-the-edge
```

**Run wrangler development server**

```bash
npx wrangler dev
```

## Resources

- [Cloudflare Wrangler docs](https://developers.cloudflare.com/workers/wrangler/get-started/)

Create a Cloudflare Worker with the Wrangler CLI

We've seen how to return easy and a little complex data from our KV, but how about in a real-world application? We are going to remove the code we have created and set up a simple HTML application that will return our data based on an API which we will create. 

Update Cloudflare Workers KVs Data from an API Route

Behind the scenes, I've updated the HTML UI and added in some helper JS functions to allow us to render new to-do's. We are going to start off by adding in the UI for a user to be able to input their own to-dos. Then we are going to work on creating a backend to store that data and retrieve it using our KV key of `"data"`. 

Filter Data with Cloudflare Workers KV Keys

By leveraging the [dependency graph](https://nx.dev/l/r/structure/dependency-graph), [Nx](https://nx.dev) is able to understand which parts in the monorepo changed and thus trigger actions on those nodes, such as building an app, running tests or linting. This already works out of the box. However, with our portfolio application, all of our markdown articles reside in the root-level `_articles` folder. Those are not captured by the Nx dependency graph and thus Nx would not correctly detect a change to our Next.js site application whenever we change or add a new markdown article.
In this lesson, we learn how to register our `_articles` folder s.t. it appears as a dedicated node in the Nx dependency graph.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/publishing-a-next-js-app-to-vercel-with-nx-df81916548f5).

Register our Markdown Articles with the Nx Dependency Graph

**NOTE:** `workspace.json` is deprecated, you can update to `project.json` [here](https://nx.dev/deprecated/workspace-json).

One cool aspect of Next.js is its [fast refresh](https://nextjs.org/docs/basic-features/fast-refresh) support. Whenever you change some Next.js component it automatically refreshes the page without hard reloading everything. However, you might notice that whenever we change something within our MDX markdown files, fast refresh is not capturing that, hence we need to manually refresh the browser window.

In this lesson, we learn how to leverage a feature of Nx to create a "[custom Next.js development server](https://nx.dev/l/r/next/server#customserverpath)". We are going to use [chokidar](https://www.npmjs.com/package/chokidar) to watch our articles folder and then invoke a private API on the Next.js server instance to trigger a reload command.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/hot-reload-mdx-changes-in-next-js-and-nx-c05252cf450a).

Add Fast Refresh Support to our MDX Files with Nx Support for Custom Next.js Servers

Next.js is a minimalistic framework for server-rendered React applications.

Next.js is all about creating performant web applications. However, the more React components we pull into our MDX via `next-mdx-remote`, the heavier our JavaScript becomes which drains our site's performance. Also, while we might have quite a few React components to be included in our Markdown rendered content, we might not always use all of them in our articles.

In this lesson, we learn how to leverage Next.js dynamic imports to just pull the minimum necessary components into our final JavaScript bundle. Also, we look into why we might want to do deep imports into our Nx library setup to decrease the size of the bundle even more.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/component-hydration-with-mdx-in-next-js-and-nx-90f46ea0431c).

Dynamically Loading React Components with next-mdx-remote and Next.js Dynamic imports

The markdown rendering includes a lot of native HTML tags as well of course. Assume for instance we want to change the behavior of links within our rendered document. We can definitely create a custom component and include that in our MDX document, but we’d rather not want to have to use a specific `<Link>` component and rather be able to keep the native way of rendering links in Markdown, that is `[some link](https://juri.dev)`.

Turns out that with `next-mdx-remote` we are able to automatically replace all rendered `<a>` tags (or really any other tag we want) with a React Component.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/component-hydration-with-mdx-in-next-js-and-nx-90f46ea0431c).

Use React Components for Native HTML Tags with MDX and Next.js

When rendering our markdown we might want to enrich the experience with some dynamic content. Many static site generators allow to include dynamic content placeholders for like embedding a Youtube video, generating table of contents, code blocks with interactive functionality and more. One way to achieve this in the React ecosystem is by using MDX. MDX is similar to JSX but for Markdown content, and allows us to directly embed React components within it. This gives an enourmous flexibility to what we can display within a page.

In this lesson we’re going to generate a new Nx library to host our “mdx elements” and learn how to leverage the `MDXRemote` component to embed these React components into our markdown rendering flow.

Embed React Components in our Markdown with MDX in a Next.js App

Once we have separated the front-matter from the rest of the markdown document, we can proceed to convert it to HTML, for then finally render it with our Next.js page component. In this lesson we use `next-mdx-remote` to parse the markdown string into HTML and we learn how to leverage the `MDXRemote` component to include it into our page component.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/read-and-render-md-files-with-next-js-and-nx-89a85c1d9b44).

Render Markdown With next-mdx-remote in our Next.js Page Component

Many markdown documents have a metadata section at the very beginning, which contains information such as the title, author and creation date. This section is the so-called "front matter". 

In order to render our markdown-based blog articles on our portfolio site we want to parse them and separate them into the front-matter and main article part. `gray-matter` is a library that helps us with that.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/read-and-render-md-files-with-next-js-and-nx-89a85c1d9b44).

Parse a Markdown Document With gray-matter

In a plain normal Next.js application you would place your markdown processing logic somewhere within a folder insider your Next application. We could do the very same in an Nx monorepo, however, we get much more benefits by moving it into a dedicated library in the `libs` folder.

[Nx workspace libraries](https://nx.dev/l/a/structure/grouping-libraries#example-workspace) allow you to nicely encapsulate your functionality as well as to easily share it with other apps that might potentially live in our workspace. In this lesson, we learn how to generate a new library with Nx generators and how to import the library into our Next.js application.

Prefer to read along as well? Here's the [accompanying article](https://blog.nrwl.io/read-and-render-md-files-with-next-js-and-nx-89a85c1d9b44).

Generate an Nx Library to Implement our Markdown Rendering for our Next.js App

Our application works but there is one major flaw. Everyone sees the same to-do's. To make to-do's user-specific, we are going to link users' data to their IP addresses so people only see to-dos based on their IP address. 

Identify Users in a Cloudflare Worker by IP to Serve User-specific Data

Now that we have our Workers project up and running, we need to create a KV namespace. A KV namespace is a key-value database that is replicated to Cloudflare's edge. Using the terminal, we will create our namespace and add it to our wrangler.toml file. 

Create a Cloudflare Workers KV Namespace

Creating a new workers project using Wrangler is pretty easy. We first will install and double-check that we have Wrangler installed. Then we will connect Wrangler to our cloudflare project. Lastly, using the terminal, we can easily create our new Wrangler project. 

Create a New Cloudflare Workers Project with Wrangler

We can store and retrieve our data from our KV but not the status of whether or not it's checked off. To do that, we create a function called `completeTodo` that will target the checkbox, find the todo based on its id, flips the boolean from false to true and saves that data, and then updates our to-dos. 

Edit Stored Data using the Cloudflare KV Key

 We map over the initial set of results then use `Promise.all` to efficiently fetch the details of each pokemon rendered in the table. This allows us to add more columns to our pokemon table.

Fetch and display row details with Promise.all in Tanstack Table

map over the initial set of results then use `Promise.all` to efficiently fetch the details of each pokemon rendered in the tanstack table

Create a custom hook to fetch remote data by passing a promise to the React 19 `use` hook. Then render this data in Tanstack Table by memoizing the results and passing them to `useReactTable`.

Render remote data in a Tanstack Table with React 19's use hook and a promise

Create a custom hook using React 19’s use to fetch data with a promise. Render it in a Tanstack Table by memoizing and using useReactTable.

You know when you're deep in the code trenches and need someone to challenge your assumptions? That's exactly what this is about.

Instead of waiting for someone to poke holes in your master plan, you can paste that strategy straight into AI Studio and let it pressure test your ideas. We're talking about using this tool to validate everything from your feature proposals to your ambitious refactoring strategies.

The real magic happens when you hit that microphone button. Now you're in a real-time verbal sparring match. It's like rubber duck debugging but the duck actually pushes back and makes you defend every architectural decision you've made.

Here's the workflow:
1. Drop your master plan into the system instructions
2. Tell the AI to challenge every assumption
3. Hit that mic and defend your choices like you mean it
4. Watch your plan either emerge stronger or get rightfully dismantled

This approach is perfect when you're:
- Stuck in your own head with a complex problem
- Need to validate if your approach holds water
- Want to catch edge cases before they catch you
- Looking for opportunities you might have missed because you're too close to the code

It's like having a senior dev on call 24/7, ready to challenge your ideas and push you to think deeper about your solutions.

Now go forth and get your ideas challenged. Your projects will thank you for it.


An Uncomfortable and Challenging Conversation with AI Studio

In this lesson, you'll learn how to effectively brainstorm your project ideas before working with AI agents. We'll cover:

1. The importance of having a clear plan before starting work with AI agents
2. Using traditional brainstorming techniques to gather your thoughts
3. Leveraging video recording to capture your brainstorming session
4. How to review and analyze websites for inspiration
5. Documenting likes, dislikes, and potential features for your project
6. Using AI tools (like AI Studio) to organize and process your recorded brainstorming session

The lesson emphasizes a modern approach to brainstorming that combines traditional techniques with video recording, making it easier to later process your ideas through AI tools. This method helps ensure that no ideas are lost and that your thoughts can be effectively organized by AI for future reference.

How to Brainstorm Your Project Features to Prepare for AI Agents

Many people share .cursorrules files, but trying to craft your own from scratch is a huge pain. Instead, this lesson shows you how to grab ALL community-created .cursorrules files to build a custom rule set for your projects. We'll use Gitingest and Google AI Studio to gather these examples and generate an awesome starting point customized to your specific project.

Generate Custom .cursorrules for Your Project Based on Community Examples

The GitHub CLI allows you to search issues and filter on different criteria like amount of comments. This allows you to search for real-world situations and pains/solutions people have using different technology. 

With the GitHub CLI in hand we'll drop 100s of issues as context in Googles AI Studio and have Gemini determine the 10 most frequently asked questions. This will give us a sense of how people use a library and where people get hung up.

Search 100s of Issues with GitHub CLI as Context for AI Studio

I often find myself wondering how to use certain APIs in certain situations, and I have a couple of keywords that I try and search up on Google. 

What you can do in this situation is turn to GitHub and see how other people have actually used this code on public projects. The GitHub CLI enables us to run searches like this, and we can take the results and work on them programmatically.

Inside of Cursor with their Composer AI agent, we can search for the files we want, tell the agent to download them, and inspect them for better understanding of how those things work together.

Search All of GitHub's Code and Pass the Results to AI Agents in Cursor

Implement virtual scrolling to handle large datasets in an HTML table. Start with a basic table that renders 10,000 rows of data. The browser can struggle with this amount of DOM nodes to render at one time, especially on low end devices. 

We implement tanstack react virtual to render only the necessary amount of rows the user is viewing. The implementation includes proper column sizing, accessibility considerations, and convenient scroll controls, resulting in a smooth and performant table experience regardless of the dataset size.

Create a row virtualized HTML table with Tanstack Virtual

Learn how to implement virtual scrolling in React tables using TanStack Table and Tanstack Virtual with the useWindowVirtualizer React hook.

The most difficult tasks will require the strongest AI models. And those models simply aren't available as AI agents inside of your tooling. So the best workflows are to drag and drop the files you need to be changed and analyzed inside of a reasoning model such as OpenAI's O1 Pro, then letting it create a giant plan for your dumber AI agents to work on. The key here is that instead of just copying and pasting the entire plan into Cursors Composer or your AI agent, it's to create a file inside of your project and ask your agent to tackle the plan one step at a time. This allows you ample opportunity between each step to test and verify the micro steps along the way. Otherwise, as I've experienced many times, a session can go way off track and you burn a lot of time and burn a lot of AI requests Cursors Composer and it starts churning through a plan Cursors Composer and making lots of unwanted changes. Cursors Composer and it's to guide an AI agent to hold its hand along the way, especially through the most complex tasks. Cursors Composer and the most complex tasks Cursors Composer and the most complex tasks require a step-by-step plan.

Planning AI Agent Sessions with Reasoning Models

AI agents will inevitably make mistakes—it's just a matter of when. A key tool in your toolbox, once they do make mistakes after working on a problem for a while, is to use **git diff** to capture all of the changes they've made. 

Once you have those changes in a text file, you have a portable document that you can test against other AI models. This allows you to move the problem into smarter reasoning models with smaller context windows, helping you isolate and focus on the issue at hand.

**git diff** has been invaluable when working with AI agents because they'll often make many good changes but also slip in one tiny error. Finding that one small mistake is **git diff**’s superpower.

Search

Search Results