Integrating with Urbit for Developers

This guide is aimed at developers who have a working knowledge of HTTP grammar (PUT, GET), and should not require any special knowledge of Urbit's internals. To learn more about a specific application within Urbit, please search for that with the docs once you have finished this guide.

Authentication

The first step in creating an application on Urbit (to be distinguished from an Urbit application) is to authenticate with the desired ship. This can be a planet, comet, or any other Urbit ship. For the purposes of this example, we will be running a "fakezod" on port 8080. Run this code, subsituting the appropriate path to your Urbit binary:

/path/to/urbit -p 8080 -F zod

(for the unfamiliar, this will create a folder named "zod" in your working directory, from which the ship will run. Use /path/to/urbit -p 8080 zod in the future)

You can run +code in the dojo to get the key, but a fakezod's key is always lidlut-tabwed-pillex-ridrup. This is our password, our API token that we will use.

Run this code in your bash terminal (requires cURL):

curl -i localhost:8080/~/login -X POST -d "password=lidlut-tabwed-pillex-ridrup"

Let's explain what this does:

You should see the following:

HTTP/1.1 200 ok
Date: Sun, 01 Jan 2020 00:00:00 GMT
Connection: keep-alive
Server: urbit/vere-0.10.8
set-cookie: urbauth-~zod=0v3.fvaqc.nnjda.vude1.vb5l6.kmjmg; Path=/; Max-Age=604800
transfer-encoding: chunked

The Date and set-cookie will be slightly different of course.

Take a look at the set-cookie line. This is now your session authentication, and will need to be sent with any future requests. It consists of three parts:

At present there is no way to separate these parts — you must pass a cookie as a cookie. This string must be in the Cookie header. The set-cookie header is only used for this request upon response. It is what browsers expect when receiving a cookie.

Interacting with Urbit

Once authenticated, you will perform all your operations on a channel. You may open as many channels as needed using the same authentication token, but this will increase complexity and memory usage on your Urbit ship.

Interacting with Urbit comprises two parts:

  1. Sending information via PUT requests, which receive HTTP status code 204 (no content) upon success. We will not receive updated information (as in a POST request).
  2. Receiving data on channels, which must first be subscribed to via an initial request. These come in the form of SSEs (Server-Sent Events)

These can be used independently, i.e. you may choose to only send information or only receive information from your Urbit, based on your application. Both can take place on the same channel.

Creating a Channel

For this example, we will use the following channel:

http://localhost:8080/~/channel/1601844290-ae45b

Let's exampine the parts. A channel is simply a URL in this pattern: {url}/~/channel/{uid} where {uid} is a string of your choosing. Most libraries default to the current unix time plus six hexadecimal characters. Your program should determine a channel for its current session and store that in its state.

The channel does not yet exist in Urbit, however. To subscribe to events on this URL, you must first make a request (PUT) to Urbit on this URL. The simplest way to do this is to hi ourselves, as if we had entered |hi our in the dojo. A hi is analogous to a ping, and does not manipulate state. Let's learn how to make requests using this as an example.

Making Requests

All requests will take the form of sending a message. Messages can take different forms, such as subscribes, acks, and pokes. These are called actions and there are more than these, but these will cover most basic use cases. You can think of them as a somewhat more sophisticated HTTP grammar. A messsage always requires an id and an action. Other fields depend on the specific action.

Let's look at the structure of a message by examining the most basic hi we will send.

{
  'id': 1, // Required. A sequential ID. Keep track of which messages you have sent.
  'action': 'poke', // Required. The action to take. poke is the most basic way of sending data, like HTTP POST
  'ship': 'zod', // Required by poke. The ship on which to perform the poke. You can only poke foreign ships with JSON, but this is the authenticated ship.
  'app': 'hood', // Required by poke. The Urbit app to which to send the data.
  'mark': 'helm-hi', // Required by poke. The "mark," or type, of data being sent.
  'json': 'Opening airlock' // Required by poke. The actual data being sent.
}

Let's open our channel by sending the following request using cURL (remember to change the cookie):

curl --header "Content-Type: application/json" \
     --cookie "urbauth-~zod=0v3.fvaqc.nnjda.vude1.vb5l6.kmjmg" \
     --request PUT \
     --data '[{"id":1,"action":"poke","ship":"zod","app":"hood","mark":"helm-hi","json":"Opening airlock"}]' \
     http://localhost:8080/~/channel/1601844290-ae45b

If you switch to your fakezod's dojo, you should see < ~zod: Opening airlock.

Note that in the JSON that we sent via --data, the payload is wrapped in []. This is because we can send multiple messages in a single payload, and Eyre expects a list of items, even if we are only sending one.

You will always receive a 204 status code if the PUT request successfully reached Urbit, authentication succeeded, and the payload was valid. This is merely an HTTP status code, however. If the request failed within Urbit itself, however, you will receive a "nack" event that you will have to process separately.

Let's look at how to receive events from Urbit.

Subscribing to Events

Now that Urbit has opened an Eyre channel at our URL, we can watch for events on that URL by sending a subscribe message. Let's look at the anatomy of a susbcribe:

{
  'id': 2, // Required. This is the next message we are sending, so we give it the next ID.
  'action': 'subscribe', // Required.
  'ship': 'zod', // Required by subscribe.
  'app': 'chat-view', // Required by subscribe. In this case, we want to watch events that occur in the chat-view app, which is responsible for handling incoming chats.
  'path': '/primary' // Required by subscribe. All apps have different path structures. In the case of chat-view, this is where we see all incoming chats.
}

This is all we need to send Urbit to say "On this URL through which I am sending information, send me SSEs whenever something occurs on this app at this path."

You should be able to send this information if you were able to send the above cURL example. However, we now need to be able to listen to the SSEs. The EventSource object is available for JavaScript in the browser context, as this is why SSEs were invented, but there should be a library for it in your chosen language. Take a look at a Browser JavaScript example:

const eventSource = new EventSource('http://localhost:8080/~/channel/1601844290-ae45b', {
  withCredentials: true // Required, sends your cookie
});

eventSource.addEventListener('message', function (event) {
  ack(Number(event.lastEventId)); // See section below
  const payload = JSON.parse(event.data); // Data is sent in JSON format
  payload.id === event.lastEventId; // The SSE spec includes event IDs. This information is duplicated in the payload.
  const data = payload.json; // Beyond this, the actual data will vary between apps
});

eventSource.addEventListener('error', function (event) {
  handleError(event);
});

Handling Received Events

Once you have received an event, it is expected that you will "ack" or "acknowledge" that you have received them. To do so, simply send a message in the following format:

{
  'id': 3, // Required. Picking up where we left off
  'action': 'ack', // Required.
  'event-id': 15, // Required by ack. Whichever SSE of which you are acknowledging receipt
}

Unsubscribing from Events

You may want to stop receiving events from a particular app path, but you don't want to close the channel altogether yet. Let's look at how to do that. Take a look at the example for subscribing. We sent that message with id: 2. We will use that ID to cancel the relevant subscription:

{
  'id': 4, // Required. We always increment the ID as we send messages.
  'action': 'unsubscribe', // Required.
  'subscription': 2 // Required by unsubscribe. This corresponds to the id of a subscribe message sent earlier.
}

Deleting a Channel

You should clean up your channel when able. To do so, simply send a delete message and Eyre will handle the rest so your program can safely exit.

{
  'id': 5, // Required.
  'action': 'delete' // Required.
}

Other Operations

There are other operations which require authentication but do not require the use of a channel. Some operations are app-specific and will not be enumerated here, but there are two operations that exist across Urbit: scry and spider.

Scrying

Scrying lets you see what an app has exposed at a particular path. It does not manipulate state.

It takes the form {url}/~/scry/{app}{path}.{mark}

Let's get the base hash of our fakezod in the same way that Landscape does it.

curl --header "Content-Type: application/json" \
     --cookie "urbauth-~zod=0v3.fvaqc.nnjda.vude1.vb5l6.kmjmg" \
     --request GET \
     http://localhost:8080/~/scry/file-server/clay/base/hash.json

Note the use of a GET request.

In this example we're scrying the file-server app at the path /clay/base/hash and using the json mark, which is most common. We receive "0" which is correct because we are at the top level of the hierarchy using a fake ship.

Scry will fail with a 500 or return a 404 if no data is found.

Running Threads with Spider

Running threads is an exception to the rule that we outlined in the section on channels. It uses a POST request and both manipulates state and receives information back. It also exposes the ability to send a sequence of commands, i.e. a "thread," hence the name.

It takes the form {url}/spider/{inputMark}/{threadname}/{outputmark}.json

This should be considered an advanced technique, but the basic structure of a request can be seen here:

curl --header "Content-Type: application/json" \
     --cookie "urbauth-~zod=0v3.fvaqc.nnjda.vude1.vb5l6.kmjmg" \
     --request POST \
     --data '[{"foo": "bar"}]' \
     http://localhost:8080/spider/graph-view-action/graph-create/json.json