Skip to main content
Version: testnet (v0.77)

Using GraphQL

Vega's GraphQL API provides read-only queries and subscriptions for getting data. Support is built into the data nodes, and some Vega data nodes may have a GraphQL interface enabled. GraphQL is an alternative to REST that provides more flexibility for building custom queries. The availability of the GraphQL API is determined by the node operator.

While REST provides multiple endpoints with small amounts of data, GraphQL provides a single endpoint that allows users to input complex queries and outputs as much information as is needed for the query.

Intro to the APIs

For general guidance and info you need to know about the APIs provided by Vega, read the Using the APIs page.

Queries​

Here is a simple query to fetch the current block height and Vega time from a data node:

{
  statistics {
    blockHeight
    vegaTime
  }
}

It would return a result in JSON as follows:

{
  "data": {
    "statistics": {
      "blockHeight": "1776332",
      "vegaTime": "2023-05-31T15:06:26.962283107Z"
    }
  }
}

Using POST​

GraphQL queries send requests to an endpoint using the POST method.

While you could use cURL:

curl ‘https://api.vega.community/graphql’ -X POST -H ‘content-type: application/json’ --data-raw ‘{"query":"{ statistics { blockHeight vegaTime }}"}’

It’s more likely that you’ll use a library for building queries, for example URQL ↗.

Using cURL is a nice way to see how libraries are actually working before choosing one.

Pagination object​

To use pagination, you will need to create a Pagination object to pass to the paginated queries. A pagination object looks like this:

{
  "first": 100,
  "after": "some-cursor",
  "last": 100,
  "before": "some-cursor"
}

You should provide at most two of four available fields. The table below describes which fields should be used together and how the GraphQL API will interpret the request.

Field CombinationPagination Results
No fields setThe first X number of results will be returned where X is the default number of records set, or all records if there are less than X
first onlyThe first X number of results will be returned where X is the number provided by first, or all records if there are less than X
first and afterThe page of X number of results that are available whose cursors are after the given cursor, or all remaining records if there are less than X
last onlyThe last X number of results will be returned where X is the number provided by last, or all records if there are less than X
last and beforeThe page of X number of results that are available whose cursors are before the given cursor, or all remaining records if there are less than X

No other combinations will work with the pagination, for example if you provide both the first and last properties, only the first will be processed and the last will be ignored.

By providing the first and after properties in the pagination object, you can traverse forward, i.e. start to end, through the paged results. Providing the last and before properties in the pagination object, you can traverse backward, i.e. end to start, through them.

Paginated result ordering​

By default, results will be returned in newest to oldest order, i.e you will always see the newest data at the top of your results. Therefore if you request the first 10 records, you will receive the 10 newest records, while requesting the last 10 records will return the 10 oldest records.

If no pagination information is provided, by default any paginated API will return the first 1000 results, or all if there are less than 1000 results.

Paginated results​

Most queries for Vega data could return a lot of results. To help navigate this, Vega’s GraphQL schema implements cursor based pagination ↗, which gives the user a unique string that can be passed back to the server to fetch data relative to the last request.

Before getting into cursor passing, let’s look at a paginated result. You can guess a result will be paginated if the query name ends with Connection.

{
  proposalsConnection(pagination: { first: 1 }) {
    pageInfo {
      hasNextPage
      endCursor
    }
    edges {
      node {
        id
        rationale {
          title
        }
      }
    }
  }
}

This would return something that looks like:

{
  "data": {
    "proposalsConnection": {
      "pageInfo": {
        "hasNextPage": true,
        "endCursor": "eyJzdGF0ZSI6NiwidmVnYV90aW1lIjoiMjAyMy0wNS0yNFQxNDowMDowMS4wMjk3MTZaIiwiaWQiOiI4NDAyNWU2ODM4N2NmNjFjMmI5MTIyOGQ3NjhkY2RkNGYxMGE5ZWU1Y2QyODI0ZmRlYTM1YjI1OTk3NmY1OWMxIn0="
      },
      "edges": [
        {
          "node": {
            "id": "84025e68387cf61c2b91228d768dcdd4f10a9ee5cd2824fdea35b259976f59c1",
            "rationale": {
              "title": "VMP-007 - Create market - LINK/USDT Future - 2023/06/30"
            }
          }
        }
      ]
    }
  }
}

There’s some added complexity here. Let’s break it down:

On the simple query, the structure of the result is basically identical to the query - so the shape is almost exactly what you've requested and nothing else.

Whereas with the paginated connections, the results contain extra nesting that isn't in the query.

This nesting includes a pageInfo property with some information about the data set. Here, you can see that with the current limit of 1, there is another page. You can also see the cursor for the last item in the result.

There's also an edges property that contains an array of edges. Each edge is made up of a node and a cursor. The data under node shows a proposal, information you would get by querying for a proposal by ID, while the cursor is a unique pointer that identifies the node of data.

After fetching this result, it's clear from the pageInfo that there are more proposals. You can use endCursor from the first result's pageInfo, or the cursor from the edge record (they should be the same) to fetch the next one, which will return the next 1 proposal, and a new cursor:

{
  proposalsConnection(pagination: { first: 1, after: "eyJzdGF0ZSI6NiwidmVnYV90aW1lIjoiMjAyMy0wNS0yNFQxNDowMDowMS4wMjk3MTZaIiwiaWQiOiI4NDAyNWU2ODM4N2NmNjFjMmI5MTIyOGQ3NjhkY2RkNGYxMGE5ZWU1Y2QyODI0ZmRlYTM1YjI1OTk3NmY1OWMxIn0="}) {
    pageInfo {
      hasNextPage
      endCursor
    }
    edges {
      node {
        id
        rationale {
          title
        }
      }
    }
  }
}

Subscriptions​

Subscriptions are like queries that keep returning data as it updates. These work similarly to the Observe endpoints in the REST service.

For example, the following query would show a result only when a new proposal came in:

subscription {
  proposals {
    id
    rationale {
      title
    }
  }
}

A more useful query might be the following one, which receives an update every time the mid-price of the specified market updates:

subscription {
  marketsData(marketIds: "2dca7baa5f7269b08d053668bca03f97f72e9a162327eebd941c54f1f9fb8f80") {
    midPrice
  }
}

Bus event subscription​

The bus event subscription is different from other subscriptions in one particular way. It has an extra filter for batch size. If you choose 0, then the subscription will deliver each event as it occurs. Otherwise, if you pick 10, for example, you'll receive results in batches of 10, but no response until there are 10 results.

Example that would provide an immediate result from the event bus query:

subscription {
  busEvents(types: [TimeUpdate], batchSize: 0) {
    event {
      ... on TimeUpdate {
        timestamp
      }
    }
  }
}
{
  "data": {
    "busEvents": [
      {
        "event": {
          "timestamp": "2023-03-05T15:05:33.177819749Z"
        }
      }
    ]
  }
}

GraphQL playground​

For a more approachable interface, GraphQL playground provides an interactive environment for you to build up queries with full integrated documentation.

For many users this will be the best way to get to grips with the schema.

Try visiting the GraphQL endpoint of a server. For example, you can use the GraphQL Playground for Fairground, the Vega operated testnet.

Note: A GraphQL playground may not be available on mainnet data nodes. It's recommended that you test queries on Fairground or by running your own data node with a GraphQL playground enabled.

Next steps​