GraphQL API Overview

We provide all services that Talk can provide via the GraphQL API documented on our GraphQL API Reference. If you’ve never heard about GraphQL before, visit http://graphql.org/ to learn the basics first.

Development

During development mode (when Talk has NODE_ENV=development) Talk will enable the GraphiQL IDE at the following route:

${ROOT_URL}api/v1/graph/iql

This is pretty powerful, as it lets you explore the API documentation on the sidebar as well as send off requests.

Making your first request

To learn a bit about how to interact with Talk, we’ll query for comments on a page of Talk. I have Talk running locally, (If you don’t and want to, checkout our Talk Quickstart).

The GraphQL endpoint we have can be used with any HTTP client available, but our examples below will use the common curl tool:

curl --request POST \
  --url http://localhost:3000/api/v1/graph/ql \
  --header 'Content-Type: application/json' \
  --data '{"query":"query GetComments($url: String!) { asset(url: $url) { title url comments { nodes { body user { username } } } }}","variables":{"url":"http://localhost:3000/"},"operationName":"GetComments"}'

When you unpack that, it’s really quite simple. We’re executing a POST request to the /api/v1/graph/ql route of the local Talk server with the GraphQL request we want to make. It’s composed of the query, variables, and operationName.

query GetComments($url: String!) {
  asset(url: $url) {
    title
    url
    comments {
      nodes {
        body
        user {
          username
        }
      }
    }
  }
}

The query itself is quite straightforward, we are grabbing the asset with the specified $url, and grabbing it’s title and the comments also (You can also look at our GraphQL API Reference for our entire schema).

We can then also specify our variables to the query being executed (in this case, the url for the page where we have comments on our local install of Talk):

{
    "url": "http://localhost:3000/"
}

It’s also sometimes common to have multiple queries within a query, which is where the operationName comes into play, where we simply specify the named query that we want to execute (in this case, GetComments).

To get a deeper understanding of GraphQL queries, read up on GraphQL Queries and Mutations.

Understanding the response

Once you completed the above GraphQL query with curl, you’ll get a response sort of like this:

{
  "data": {
    "asset": {
      "title": "Coral Talk",
      "url": "http://localhost:3000/",
      "comments": {
        "nodes": [
          {
            "body": "Second comment!",
            "user": {
              "username": "wyattjoh"
            }
          },
          {
            "body": "First comment!",
            "user": {
              "username": "wyattjoh"
            }
          }
        ]
      }
    }
  }
}

All of the parameters you requested should be available under the data property. Any errors that you get would appear in a errors array at the top level, like this:

{
  "data": {
    "asset": null
  },
  "errors": [
    {
      "message": "asset_url is invalid",
      "locations": [
        {
          "line": 2,
          "column": 3
        }
      ],
      "path": [
        "asset"
      ]
    }
  ]
}

You should know that any property that is marked with a ! is considered required, and non-nullable, which means you can always guarantee on it being there in your request if there were no errors.

Authorizing a request

Some queries you may notice seem to return null or an error of NOT_AUTHORIZED. It’s likely the case that you are making a request to a route that requires authorization. You can perform authorization a few ways in Talk:

1. As a Bearer Token
2. As a Query Parameter
3. As a Cookie

Essentially, you need to get access to a JWT token that you can use to authorize your requests. Generating one is simple, you can use the CLI tools in Talk to do that.

# first, find your user account
./bin/cli users list

# then, create a token for that account
./bin/cli token create ${USER_ID} example-token

Where USER_ID is the ID of your user account you found using the users list command.

Once you have your access token, you can substitute it as ${TOKEN} in your curl request as follows:

Bearer Token

curl --request POST \
  --url http://localhost:3000/api/v1/graph/ql \
  --header 'Content-Type: application/json' \
  --header "Authorization: Bearer ${TOKEN}"
  --data '{"query":"query GetComments($url: String!) { asset(url: $url) { title url comments { nodes { body user { username } } } }}","variables":{"url":"http://localhost:3000/"},"operationName":"GetComments"}'

Query Parameter

curl --request POST \
  --url http://localhost:3000/api/v1/graph/ql?access_token=${TOKEN} \
  --header 'Content-Type: application/json' \
  --data '{"query":"query GetComments($url: String!) { asset(url: $url) { title url comments { nodes { body user { username } } } }}","variables":{"url":"http://localhost:3000/"},"operationName":"GetComments"}'

Cookie

curl --request POST \
  --url http://localhost:3000/api/v1/graph/ql \
  --header 'Content-Type: application/json' \
  --cookie "authorization=${TOKEN}"
  --data '{"query":"query GetComments($url: String!) { asset(url: $url) { title url comments { nodes { body user { username } } } }}","variables":{"url":"http://localhost:3000/"},"operationName":"GetComments"}'