Zero to Built: Creating a Forum App

Why do we use the term DNA?

With Holochain, we’re trying to enable people to choose to participate in coherent social coordination, or interact meaningfully with each other online without needing a central authority to define the rules and keep everyone safe. To do that, we are borrowing some patterns from how biological organisms are able to coordinate coherently even at scales that social organizations such as companies or nations have come nowhere close to. In living creatures like humans, dolphins, redwood trees, and coral reefs, many of the cells in the body of an organism (trillions of the cells in a human body, for instance) are each running a (roughly) identical copy of a rule set in the form of DNA.

This enables many different independent parts (cells) to build relatively consistent superstructures (a body, for instance), move resources, identify and eliminate infections, and more — all without centralized command and control. There is no “CEO” cell in the body telling everybody else what to do. It’s a bunch of independent actors (cells) playing by a consistent set of rules (the DNA) coordinating in effective and resilient ways.

A cell in the muscle of your bicep finds itself in a particular context, surrounded by certain resources and conditions. Based on those signals, that cell behaves in particular ways, running relevant portions of the larger shared instruction set (DNA) and transforming resources in ways that make sense for a bicep muscle cell. A different cell in your blood, perhaps facing a context where there is a bacterial infection, will face a different set of circumstances and consequently will make use of other parts of the shared instruction set to guide how it behaves. In other words, many biological organisms make use of this pattern where all participants run the same rule set, but each in its own context, and that unlocks a powerful capacity for coherent coordination.

Holochain borrows this pattern that we see in biological coordination to try to enable similarly coherent social coordination. However, our focus is on enabling participants to “opt into” such coherent social coordination. We believe that this pattern of being able to choose which games you want to play — and being able to leave them or adapt them as experience suggests — is critical to enabling individual and collective adaptive capacity. We believe that it may enable a fundamental shift in the ability of individuals and communities to sense and respond to the situations that they face.

To put it another way: if a group of us can all agree to the rules of a game, then together we can play that game.

All of us opting in to those rules — and helping to enforce them — enables us to play that game together, whether it is a game of chess, chat, a forum app, or something much richer.

DNA as boundary of network

The network of participants that are running a DNA engage in “peer witnessing” of each other’s actions in that network. A (deterministically random) set of peers are responsible for validating, storing, and serving each particular piece of shared content. In other words, the users of a particular hApp agree to a set of rules and then work together collectively to enforce those rules and to store and serve content (state changes) that do not violate those rules.

Every hApp needs to include at least one DNA. Moreover, as indicated above, it is at the DNA level (note: not the higher application level) where participants will form a network of peers to validate, store, and serve content in accordance with the rules defined in that DNA. This happens in the background as the application runs on each participant’s machine.

There are some powerful consequences to this architectural choice — including freedom to have your application look and feel the way you want, or to combine multiple DNAs together in ways that work for you without having to get everyone else to agree to do the same — but we’ll save those details for later.

So if we have multiple DNAs in our hApp…

…then we are participating in multiple networks, with each network of peers that are participating in a particular DNA also helping maintain the shared database for that DNA, enforcing its rules while validating, storing, and serving content. Each network acts as a ‘social organism’ in cooperation with other networks in the hApp.

This is similar to the way in which multiple DNA communities coexist in biological organisms. In fact, there are more cells in a human body that contain other DNA (like bacteria and other microorganisms) than cells that contain our DNA. This indicates that we are an ecology of coherent communities that are interacting with — and evolving alongside — one another.

When it comes to hApps, this lets us play coherent games with one another at the DNA level, while also participating in adjacent coherent games with others as well. That means that applications are not one-size-fits-all. You can choose to combine different bits of functionality in interesting and novel ways.

Integrity zomes

An integrity zome, as the name suggests, is responsible for maintaining the data integrity of a Holochain application. It sets the rules and ensures that any data writes occurring within the application are consistent with those rules. In other words, it’s responsible for ensuring that data is correct, complete, and trustworthy. Integrity zomes help maintain a secure and reliable distributed peer-to-peer network by enforcing the validation rules defined by the application developer — in this case, you!

Coordinator zomes

On the other hand, a coordinator zome contains the code that actually commits data, retrieves it, or sends and receives messages between peers or between other portions of the application on a user’s own device (between the back end and the UI, for instance). A coordinator zome is where you define the API for your DNA, through which the network of peers and their data is made accessible to the user.

Multiple zomes per DNA

As you learned earlier, a DNA can have multiple integrity and coordinator zomes. Each integrity zome contributes to the full set of different types of valid data that can be written, while each coordinator zome contributes to the DNA’s functionality that you expose through its API. In order to write data of a certain type, a coordinator zome needs to specify a dependency on the integrity zome that defines that data type. A coordinator zome can also depend on multiple integrity zomes.

Why two types?

They are separated from one another so we can update coordinator zomes without having to update the integrity zomes. This is important, because changes made to an integrity zome result in a change of the rule set, which results in an entirely new network. This is because the integrity code is what defines the ‘rules of the game’ for a group of participants. If you changed the code of an integrity zome, Holochain would consider it a new ‘game’ and you would find yourself suddenly in a new and different network from the other folks who haven’t yet changed their integrity zome — and we want to minimize those sorts of forks to situations where they are needed (like when a community decides they want to play by different rules, for instance changing the maximum length of comments from 140 characters to 280 characters).

At the same time, a community will want to be able to improve the ways in which things are done in a Holochain app. This can take the form of adding new features or fixing bugs, and we also want people to also be able to take advantage of the latest features in Holochain. Separating integrity and coordination enables them to do that more easily, because:

• Holochain’s coordinator zome API receives frequent updates while the integrity zome API is fairly stable, and
• coordinator zomes can be added to or removed from a DNA at runtime without affecting the DNA’s hash.

Source chain

Any time a participant in a hApp takes some action that creates or changes data, they do it by adding a record to a journal called a source chain. Each participant has their own source chain, a local, tamper-proof, and chronological store of the participant’s actions in that application.

This is one of the main differences between Holochain and other systems such as blockchains or centralized server-based applications. Instead of recording a “global” (community-wide) timeline of what actions have taken place, in Holochain actions are recorded on individual timelines. They can be thought of as both a change to the individual’s state and an attempt to change shared state.

One big advantage of this approach is that a single agent can be considered authoritative about the order in which they took actions. From their perspective, first they did A, then B, then C, etc. The fact that someone else didn’t get an update about these changes, and possibly received them in a different order, doesn’t matter. The order that the authoring agent took those actions will be captured in the actions themselves (thanks to each action referencing the previous one that they had taken, thus creating an ordered sequence — or chain — of actions).

Actions and entries

You’ll notice that we used the word “action” a lot. In fact, we call an item on the source chain an action. In Holochain applications, data is always “spoken into being” by an agent (a participant). Each action captures their act of adding, modifying, or removing data, rather than simply capturing the data itself.

There are a few different kinds of actions, but the most common one is Create, which creates an ‘entry’ — an arbitrary blob of bytes. Entries store most of the actual content created by a participant, such as the text of a post in our forum hApp. When someone creates a forum post, they’re recording an action to their source chain that reads something like: I am creating this forum post entry with the title “Intros” and the content “Where are you from and what is something you love about where you live?” and I would like my peers in the network to publicly store a record of this act along with the data itself. So while an action is useful for storing noun-like data like messages and images, it’s actually a verb, a record of an action that someone took to update their own state and possibly the shared state as well. That also makes it well-suited to verb-like data like document edits, game moves, and transactions.

Every action contains the ID of its author (actually a cryptographic public key), a timestamp, a pointer to the previous source chain record, and a pointer to the entry data, if there is any. In this way, actions provide historical context and provenance for the entries they operate on.

The pointer to the previous source chain record creates an unbroken history from the current record all the way back to the source chain’s starting point. This ‘genesis’ record contains the hash of the DNA, which serves as both the identifier for the specific set of validation rules that all following records should follow and the ID of the network that this source chain’s actions are participating in.

An action is cryptographically signed by its author and is immutable (can’t be changed or erased from either the source chain or the network’s data store) once written. This, along with the validation rules specified by the DNA hash in the genesis record, are examples of a concept we call “intrinsic data integrity”, in which data carries enough information about itself to be self-validating.

Just as with a centralized application, we aren’t just going to add this data into some database without checking it first. When a participant tries to write an action, Holochain first:

1. ensures that the action being taken doesn’t violate the validation rules of the DNA,
2. adds it as the next record to the source chain, and then
3. tells the participant’s network peers about it so they can validate and store it, if it’s meant to be public.

The bits of shared information that all the peers in a network are holding are collectively called a distributed hash table, or DHT. We’ll explain more about the DHT later.

If you want to learn more, check out The Source Chain: A Personal Data Journal and The DHT: A Shared, Distributed Graph Database. You’ll also get to see it all in action in a later step, when you run your hApp for the first time.

An entry type is just a label, an identifier for a certain type of data that your DNA deals with. It serves as something to attach validation rules to in your integrity zome, and those rules are what give an entry type its meaning. They take the form of code in a function that gets called any time something is about to be stored, and because they’re just code, they can validate all sorts of things. Here are a few key examples:

• Data structure: When you use the scaffolding tool to create an entry type, it generates a Rust struct that defines fields in your entry type, and it also generates code in the validation function that attempts to convert the raw bytes into an instance of that type. By providing a well-defined structure, this type ensures that data can be understood by the application. If it can’t be deserialized into the appropriate Rust structure, it’s not valid.

• Constraints on data: Beyond simple fields, validation code can constrain the values in an entry — for instance, it can enforce a maximum number of characters in a text field or reject nonsensical calendar dates.

• Privileges: Because it originates in a source chain, an entry comes with metadata about its author. This can be used to control who can create, edit, or delete an entry.

• Context: Because an action is part of a chain of actions, it can be validated based on the agent’s history — for instance, to prevent currency transactions beyond a credit limit or disallow more than two comments per minute to discourage spam. An entry can also point to other actions and entries in the DHT upon which it depends, and their data can be used in its validation.

Mutating immutable data and improving performance

In short, the above choice is about how changes get dealt with when a piece of content is updated.

Because all data in a Holochain application is immutable once it’s written, we don’t just go changing existing content, because that would break the integrity of the agent’s source chain as well as the data already in the DHT. So instead we add metadata to the original data, indicating that people should now look elsewhere for the data or consider it deleted. This is produced by Update and Delete source chain actions.

For an Update action, the original Create or Update action and its entry content on the DHT get a ReplacedBy pointer to the new Update action and its entry content.

When the scaffolding tool asks you whether to create a link from the original entry, though, it’s not talking about this pointer. Instead, it’s talking about an extra piece of metadata that points to the very newest entry in a chain of updates. If an entry were to get updated, and that update were updated, and this were repeated three more times, anyone trying to retrieve the entry would have to query the DHT six times before they finally found the newest revision. This extra link, which is not a built-in feature of updates, ‘jumps’ them past the entire chain of updates at the cost of a bit of extra storage. The scaffolding tool will generate all the extra code needed to write and read this metadata in its update and read functions. You can read more about links in the Core Concepts section on Links and Anchors.

For a Delete action, the original action and its entry content are simply marked as dead. In the cases of both updating and deleting, all original data is still accessible if the application needs it.

Resolving conflicts

Multiple participants can mark a single entry as updated or deleted at the same time. This might be surprising, but Holochain does this for two good reasons. First, it’s surprisingly difficult to decide which is the ‘correct’ version of a piece of data in a distributed system, because contributions may come from any peer at any time, even appearing unexpectedly long after they’ve been created. There are many strategies for resolving the conflicts that arise from this, which brings us to the second good reason: we don’t want to impose a specific conflict resolution strategy on you. Your application may not even consider parallel updates and deletes on a single entry to be a conflict at all.

CRUD functions

By default, the scaffolding tool generates a `create_<entry_type>’ function in your coordinator zome for an entry type because creating new data is a fundamental part of any application, and it reflects the core principle of Holochain’s agent-centric approach — the ability to make changes to your own application’s state.

Similarly, when a public entry is published, it becomes accessible to other agents in the network. Public entries are meant to be shared and discovered by others, so a `read_<entry_type>’ function is provided by default to ensure that agents can easily access and retrieve publicly shared entries. (The content of private entries, however, are not shared to the network.) For more info on entries, see: the Core Concepts sections on Source Chains and DHT.

Developers decide whether to let the scaffolding tool generate update_<entry_type> and delete_<entry_type> functions based on their specific application requirements. More details in the Core Concepts section on CRUD.

There are two kinds of unique identifiers or ‘addresses’ in Holochain: hashes for data and public keys for agents.

A hash is a unique “digital fingerprint” for a piece of data, generated by running it through a mathematical function called a hash function. None of the original data is present in the hash, but even so, the hash is extremely unlikely to be identical to the hash of any other piece of data. If you change even one character of the entry’s content, the hash will be radically (and unpredictably) different.

Holochain uses a hash function called blake2b. You can play with an online blake2b hash generator to see how changing content a tiny bit alters the hash. Try hashing hi and then Hi and compare their hashes.

To ensure data integrity and facilitate efficient data retrieval, each piece of data is identified by its hash. This serves the following purposes:

• Uniqueness: The cryptographic hashing function ensures that the data has a unique hash value, which helps to differentiate it from other data on the network.
• Efficient lookup: The hash is used as a key (essentially an address) in the network’s storage system, the distributed hash table (DHT). When an agent wants to retrieve data, they simply search for it by hash, without needing to know what peer machine it’s stored on. In the background, Holochain reaches out simultaneously to multiple peers who are responsible for the hash based on an algorithm that matches peers to data based on the similarity of the hash to their agent IDs. This makes data lookup fast and resilient to unreliable peers or network conditions.
• Fair distribution: Because the participants in a network are responsible for validating and storing each other’s public data based on its hash, the randomness of the hashing function ensures that the responsibility for the entire data set is spread fairly evenly among everyone.
• Integrity verification: Hi will always generate the same hash no matter who runs it through the hashing function. So when data is retrieved by hash, its hash can be recalculated and compared with the original requested hash to ensure that a third party hasn’t tampered with the data.
• Collusion resistance: The network peers who take responsibility for validating and storing an entry are chosen randomly based on the similarity of their agent IDs to the EntryHash. It would take a huge amount of computing power to generate a hash that would fall under the responsibility of a colluding peer. And because Holochain can retrieve data from multiple peers, it’s more likely that the requestor can find one honest peer to report problems with a piece of bad data.

ActionHash

An action is identified by its ActionHash. Because an action contains information about its author, the time it was written, the action that preceded it, and the entry it operates on, no two action hashes will be the same — even for two Create actions that write the same entry. This helps to disambiguate identical entries written at different times or by different agents.

Because of the built-in fields and the uniqueness of each action hash, whenever you want to reference another piece of data, the most sensible choice is often an ActionHash.

EntryHash

An entry is identified by its EntryHash, which can be retrieved from the action that wrote it. Because they’re two separate pieces of data, an entry and its action are stored by different peers in the network.

More than one agent (and in fact more than one action from the same agent) can write the same entry, so entry hashes aren’t guaranteed to be unique. Sometimes this is a problem, but sometimes it’s what you want when you’re referencing data – for instance, when the content you want to refer to is more important than who authored it.

AgentPubKey

Each agent in a network is identified by their cryptographic public key, a unique number that’s mathematically related to a private number that they hold on their machine. Public-key cryptography is a little complex for this guide — it’s enough to know that a participant’s private key signs their source chain actions, and those signatures paired with their public key allow others to verify that they are the one who authored those actions.

An AgentPubKey isn’t a hash, but it’s the same length as one, and it’s unique just like a hash. So it can be used as a way of referring to an agent, like a user ID — and this is also why it’s used to choose peers in the DHT storage and retrieval algorithm.

Summary

Whereas EntryHash is used to uniquely identify, store, and efficiently retrieve an entry from the DHT, ActionHash does the same for the action that operated on it, which can provide information about the history and context of any associated entry (including what action preceded it). ActionHashes are also what enable any participant to retrieve and reconstruct the continuous sequence of actions (and any associated entries) in another agent’s source chain.

Use EntryHash when you want to link to or retrieve the actual content or data (e.g., when linking to a category in a forum application).

Use ActionHash when you want to link to or retrieve the entry content along with its authorship context (e.g., when distinguishing between two posts with identical content).

Use AgentPubKey when you want to link to an agent (such as associating a profile or icon with them) or retrieve information about their history (such as scanning their source chain for posts and comments).

You can check out the Core Concepts to dive a bit deeper into how the distributed hash table helps to not only make these entries and actions available but helps to ensure that agents haven’t gone back to try and change their own histories after the fact. But for now, let’s dive into links.

What exactly is a link? Where is it stored? How does it work? And what do they let us do that we couldn’t do otherwise?

Links enable us to build a graph of references from one piece of content to other pieces of content in a DHT and then to navigate that graph. This is important because without some sort of trail to follow, it is infeasible to just “search for all content” thanks to the address space (all possible hashes) being so large and spread out across machines that iterating through tme all could take millions of years.

By linking from known things to unknown things, we enable the efficient discovery and retrieval of related content in our hApp.

Storage: When an agent creates a link between two entries, a CreateLink action is written to their source chain. A link is so small that there’s no entry for the action. It simply contains the address of the base, the address of the target, the link type (which describes the relationship), and an optional tag which contains a small amount of application-specific information. The base and target can be any sort of DHT address — an EntryHash, an ActionHash, or an AgentPubKey. But there doesn’t actually need to be any data at that base, which is useful for referencing data that exists in another hash-based data store outside the DHT.

After storing the action in the local source chain, the agent then publishes the link to the DHT, where it goes to the peers who are responsible for storing the base address and gets attached to the address as metadata.

Lookup: To look up and retrieve links in a Holochain app, agents can perform a get_links query on a base DHT address. This operation involves asking the DHT peers responsible for that address for any link metadata of a given link type attached to it, with an optional “starts-with” query on the link tag. The peers return a list of links matching the query, which contain the addresses of the targets, and the link types and tags. The agent can then retrieve the actual target data by performing a get query on the target address, which may be an EntryHash, ActionHash, or AgentPubKey (or an empty result, in the case of data that doesn’t exist on the DHT).

For more information and examples, read the Core Concepts section on Links and Anchors.

We already explored how links make data in the DHT discoverable by connecting known DHT base addresses to unknown addresses. Essentially every address becomes an anchor point to hang a collection of links from.

But there’s one remaining problem: where do you start? When someone starts their app for the first time, the only DHT base addresses they know about are their public key, the DNA hash, and the few actions and entries on their source chain. There’s no obvious way to start discovering other people’s data yet.

This is where collections help out. A collection is just a bunch of links on a base address that’s easy to find — typically the address is hard-coded in the coordinator zome’s code as the hash of a string such as "all_posts". It’s easy to get the links, because their base address is right there in the code.

This pattern, which we call the “anchor pattern”, is so useful that it’s built right into Holochain’s SDK — integrity zomes that use it will have all the necessary entry and link types automatically defined, and coordinator zomes that use it will have functions that can retrieve anchors and the links attached to them. The scaffolded code uses this implementation behind the scenes.

The built-in implementation is actually a simplification of a more general pattern called “paths”, which is also built into the SDK. With paths, you can create trees of linked anchors, allowing you to create and query hierarchical structures. This can be used to implement categories, granular collections (for example, “all posts” → “all posts created in 2023” → “all posts created in 2023-05” → “all posts created on 2023-05-30”), and indexes for type-ahead search (for example, “all usernames” → “all usernames starting with ‘mat’” → “all usernames starting with ‘matt’”). What the SDK calls an Anchor is actually a tree with a depth of two, in which the root node is two empty bytes.

Hierarchical paths serve another useful purpose. On the DHT, where every node is tasked with storing a portion of the whole data set, some anchors could become “hot spots” — that is, they could have thousands or even millions of links attached to them. The nodes responsible for storing those links would bear a disproportionate data storage and serving burden.

The examples of granular collections and type-ahead search indexes breaks up those anchors into increasingly smaller branches, so that each leaf node in the tree — and hence each peer — only has to store a small number of links.

The scaffolding tool doesn’t have any feature for building anchors and trees beyond simple one-anchor collections, but if you’d like to know more, you can read the Core Concepts section on Links and Anchors and the SDK reference for hash_path and anchor.

<script> section

<script lang="ts">
import type { ActionHash, AppClient } from "@holochain/client";
import { AppWebsocket } from "@holochain/client";
import { onMount, setContext } from "svelte";
import "@material/mwc-circular-progress";

import { clientContext } from "./contexts";

let client: AppClient | undefined;

let loading = true;

onMount(async () => {
  client = await AppWebsocket.connect();

  loading = false;
});

setContext(clientContext, {
  getClient: () => client,
});
</script>

This section contains the JavaScript/TypeScript code for the component. It imports various dependencies needed to build a single-page web app:

• svelte is the Svelte engine itself, and its onMount function lets you register a handler to be run when the component is initialized, while setContext lets you pass data to be used in the rendering of the component.
• @holochain/client is the Holochain client library; first we load in some useful Holochain-related TypeScript types, followed by the client object itself.
• @mwc/material-circular-progress is just a UI component that gives us a spinner when something is loading.
• ./contexts is generated by the scaffolding tool. It just contains a constant, the app-wide name for the ‘context’ that makes the Holochain client accessible to all components. In Svelte, a context is a state shared across components.

After importing dependencies, it does some initial setup. This is run when the component file is imported — in this case the component, App.svelte, is the main component for the entire application, and it’s imported into main.ts where it’s ‘mounted’ (more on mounting in a moment).

Next some variables are instantiated: one to hold the Holochain client that connects to the hApp backend via the conductor, and one to keep track of whether the client is connected yet. (This variable will be used to show a loading spinner while it’s still connecting.)

Finally, there’s an onMount handler, which is run when the component is first displayed. The handler currently does one thing: it connects to the hApp backend via the conductor, waits until the connection is established, sets loading to false, and adds the resulting client connection to the context so that all components can access it.

<main> section

<main>
  {#if loading}
    <div
      style="display: flex; flex: 1; align-items: center; justify-content: center"
    >
      <mwc-circular-progress indeterminate />
    </div>
  {:else}
    <div id="content" style="display: flex; flex-direction: column; flex: 1;">
      <h2>EDIT ME! Add the components of your app here.</h2>

      <span>Look in the <code>ui/src/DNA/ZOME</code> folders for UI elements
        that are generated with <code>hc scaffold entry-type</code>, <code>hc
          scaffold collection</code> and <code>hc scaffold link-type</code> and
        add them here as appropriate.</span>

      <span>For example, if you have scaffolded a "todos" dna, a "todos" zome, a
        "todo_item" entry type, and a collection called "all_todos", you might
        want to add an element here to create and list your todo items, with the
        generated <code>ui/src/todos/todos/AllTodos.svelte</code> and <code
        >ui/src/todos/todos/CreateTodo.svelte</code> elements.</span>

      <span>So, to use those elements here:</span>
      <ol>
        <li>
          Import the elements with:
          <pre
          >
import AllTodos from './todos/todos/AllTodos.svelte';
import CreateTodo from './todos/todos/CreateTodo.svelte';
        </pre
          >
        </li>
        <li>
          Replace this "EDIT ME!" section with <code
          >&lt;CreateTodo&gt;&lt;/CreateTodo&gt;&lt;AllTodos&gt;&lt;/AllTodos&gt;</code
          >.
        </li>
      </ol>
    </div>
  {/if}
</main>

This section is a template for the displayable content of the main app component. Using an {#if} block to test whether the reactive variable loading is true, this section displays a spinner until the backend can be accessed. Once the UI is connected to the backend, it shows some boilerplate text telling you to add something meaningful to the template.

Note that, in Svelte, any time a variable changes, the template is re-rendered with the new value. This is called reactivity, and makes your life easier because you don’t have to write quite so many event handlers for changes on your data.

<style> section

<style>
  main {
    text-align: center;
    padding: 1em;
    max-width: 240px;
    margin: 0 auto;
  }

  @media (min-width: 640px) {
    main {
      max-width: none;
    }
  }
</style>

This section is a template for the CSS styles that get applied to the HTML in the <main> section of the component. You can also use reactive variables here, and the styling will update whenever the variables change. These scaffolded styles set the component up with some basic layout to make it readable at small and large window sizes.

All Svelte components follow this general pattern. App.svelte has special status as the root component, but otherwise it’s just like any other component.

At this point, in our DHT graph it should look like we have two different agents and then a separate floating entry and action. But we know that the new post is associated with a source chain which is associated with an agent. So why aren’t they connected on the DHT?

A source chain merely serves as a history of one agent’s attempts to manipulate the state of the graph database contained in the DHT. It’s useful to think of the DHT as a completely separate data store that doesn’t necessarily reflect agent-to-entry relationships unless you explicitly create a link type for them.

For the purpose of this hApp, we’re not interested in agent-to-posts relationships, so it’s fine that they’re not linked. But if you wanted to create a page that showed all posts by an author, that’s when you might want to scaffold that link type. hc scaffold collection will do this for you if you choose a by-author collection, and will also create a corresponding coordinator function and Svelte component.