Rate limits and query limits for the GraphQL API

Primary rate limit

The GraphQL API assigns points to each query and limits the points that you can use within a specific amount of time. This limit helps prevent abuse and denial-of-service attacks, and ensures that the API remains available for all users.

The REST API also has a separate primary rate limit. For more information, see REST API 的速率限制.

In general, you can calculate your primary rate limit for the GraphQL API based on your method of authentication:

For users: 5,000 points per hour per user. This includes requests made with a personal access token as well as requests made by a GitHub App or OAuth app on behalf of a user that authorized the app. Requests made on a user's behalf by a GitHub App that is owned by a GitHub Enterprise Cloud organization have a higher rate limit of 10,000 points per hour. Similarly, requests made on your behalf by an OAuth app that is owned or approved by a GitHub Enterprise Cloud organization have a higher rate limit of 10,000 points per hour if you are a member of the GitHub Enterprise Cloud organization.
For GitHub App installations not on a GitHub Enterprise Cloud organization: 5,000 points per hour per installation. Installations that have more than 20 repositories receive another 50 points per hour for each repository. Installations that are on an organization that have more than 20 users receive another 50 points per hour for each user. The rate limit cannot increase beyond 12,500 points per hour. The rate limit for user access tokens (as opposed to installation access tokens) are dictated by the primary rate limit for users.
For GitHub App installations on a GitHub Enterprise Cloud organization: 10,000 points per hour per installation. The rate limit for user access tokens (as opposed to installation access tokens) are dictated by the primary rate limit for users.
For OAuth apps: 5,000 points per hour, or 10,000 points per hour if the app is owned by a GitHub Enterprise Cloud organization. This only applies when the app uses their client ID and client secret to request public data. The rate limit for OAuth access tokens generated by a OAuth app are dictated by the primary rate limit for users.
For GITHUB_TOKEN in GitHub Actions workflows: 1,000 points per hour per repository. For requests to resources that belong to an enterprise account on GitHub.com, the limit is 15,000 points per hour per repository.

You can check the point value of a query or calculate the expected point value as described in the following sections. The formula for calculating points and the rate limit are subject to change.

Checking the status of your primary rate limit

You can use the headers that are sent with each response to determine the current status of your primary rate limit.

Header name	Description
`x-ratelimit-limit`	The maximum number of points that you can use per hour
`x-ratelimit-remaining`	The number of points remaining in the current rate limit window
`x-ratelimit-used`	The number of points you have used in the current rate limit window
`x-ratelimit-reset`	The time at which the current rate limit window resets, in UTC epoch seconds
`x-ratelimit-resource`	The rate limit resource that the request counted against. For GraphQL requests, this will always be `graphql`.

You can also query the rateLimit object to check your rate limit. When possible, you should use the rate limit response headers instead of querying the API to check your rate limit.

query {
  viewer {
    login
  }
  rateLimit {
    limit
    remaining
    used
    resetAt
  }
}

Field	Description
`limit`	The maximum number of points that you can use per hour
`remaining`	The number of points remaining in the current rate limit window
`used`	The number of points you have used in the current rate limit window
`resetAt`	The time at which the current rate limit window resets, in UTC epoch seconds

Returning the point value of a query

You can return the point value of a query by querying the cost field on the rateLimit object:

query {
  viewer {
    login
  }
  rateLimit {
    cost
  }
}

Predicting the point value of a query

You can also roughly calculate the point value of a query before you make the query.

Add up the number of requests needed to fulfill each unique connection in the call. Assume every request will reach the first or last argument limits.
Divide the number by 100 and round the result to the nearest whole number to get the final aggregate point value. This step normalizes large numbers.

注意

The minimum point value of a call to the GraphQL API is 1.

Here's an example query and score calculation:

query {
  viewer {
    login
    repositories(first: 100) {
      edges {
        node {
          id

          issues(first: 50) {
            edges {
              node {
                id

                labels(first: 60) {
                  edges {
                    node {
                      id
                      name
                    }
                  }
                }
              }
            }
          }
        }
      }
    }
  }
}

This query requires 5,101 requests to fulfill:

Although we're returning 100 repositories, the API has to connect to the viewer's account once to get the list of repositories. So, requests for repositories = 1
Although we're returning 50 issues, the API has to connect to each of the 100 repositories to get the list of issues. So, requests for issues = 100
Although we're returning 60 labels, the API has to connect to each of the 5,000 potential total issues to get the list of labels. So, requests for labels = 5,000
Total = 5,101

Dividing by 100 and rounding gives us the final score of the query: 51

Secondary rate limits

除了主要速率限制以外，GitHub 还强制执行次要速率限制以阻止滥用，让 API 可供所有用户所使用。

以下情况下可能会遇到次要速率限制：

发出的并发请求过多。 并发请求数量不能超过 100 个。 REST API 和 GraphQL API 都应用此限制。
每分钟向单个终结点发出的请求数过多。 REST API 终结点每分钟允许发出的请求数不超过 900 点，GraphQL API 终结点每分钟允许发出的请求数不过超 2,000 点。有关点的详细信息，请参阅“计算次要速率限制的点数”。
每分钟发出的请求数过多。 实时每 60 秒允许的 CPU 时间不超过 90 秒。此 CPU 时间的 60 秒不能超过 GraphQL API。可以通过衡量 API 请求的总响应时间来大致估算出 CPU 时间。
发出过多的请求，它们在短时间内会消耗过多的计算资源。
短时间内在 GitHub 上创建的内容过多。 一般情况下，每分钟不超过 80 个内容生成请求，允许每小时不超过 500 个内容生成请求。某些终结点的内容创建限制较低。内容创建限制包括对 GitHub Web 界面以及 REST API 和 GraphQL API 执行的操作。

上述次要速率限制可能随时更改，恕不另行通知。可能还会因未公开的原因而遇到次要速率限制。

计算次要速率限制的点数

某些次要速率限制由请求的点值确定。对于 GraphQL 请求，这些点值与主要速率限制的点值分开来进行计算。

请求	点
不具有突变的 GraphQL 请求	1
具有突变的 GraphQL 请求	5
大多数 REST API `GET`、`HEAD` 和 `OPTIONS` 请求	1
大多数 REST API `POST`、`PATCH`、`PUT` 或 `DELETE` 请求	5

某些 REST API 终结点具有不公开共享的不同点成本。

Exceeding the rate limit

If you exceed your primary rate limit, the response status will still be 200, but you will receive an error message, and the value of the x-ratelimit-remaining header will be 0. You should not retry your request until after the time specified by the x-ratelimit-reset header.

If you exceed a secondary rate limit, the response status will be 200 or 403, and you will receive an error message that indicates that you hit a secondary rate limit. If the retry-after response header is present, you should not retry your request until after that many seconds has elapsed. If the x-ratelimit-remaining header is 0, you should not retry your request until after the time, in UTC epoch seconds, specified by the x-ratelimit-reset header. Otherwise, wait for at least one minute before retrying. If your request continues to fail due to a secondary rate limit, wait for an exponentially increasing amount of time between retries, and throw an error after a specific number of retries.

Continuing to make requests while you are rate limited may result in the banning of your integration.

Staying under the rate limit

To avoid exceeding a rate limit, you should pause at least 1 second between mutative requests and avoid concurrent requests.

You should also subscribe to webhook events instead of polling the API for data. For more information, see Webhook 文档.

You can also stream the audit log in order to view API requests. This can help you troubleshoot integrations that are exceeding the rate limit. For more information, see 流式处理企业审核日志.

Node limit

To pass schema validation, all GraphQL API calls must meet these standards:

Clients must supply a first or last argument on any connection.
Values of first and last must be within 1-100.
Individual calls cannot request more than 500,000 total nodes.

Calculating nodes in a call

These two examples show how to calculate the total nodes in a call.

Simple query:

query {
  viewer {
    repositories(first: 50) {
      edges {
        repository:node {
          name

          issues(first: 10) {
            totalCount
            edges {
              node {
                title
                bodyHTML
              }
            }
          }
        }
      }
    }
  }
}

Calculation:

50         = 50 repositories
 +
50 x 10  = 500 repository issues

            = 550 total nodes

Complex query:

query {
  viewer {
    repositories(first: 50) {
      edges {
        repository:node {
          name

          pullRequests(first: 20) {
            edges {
              pullRequest:node {
                title

                comments(first: 10) {
                  edges {
                    comment:node {
                      bodyHTML
                    }
                  }
                }
              }
            }
          }

          issues(first: 20) {
            totalCount
            edges {
              issue:node {
                title
                bodyHTML

                comments(first: 10) {
                  edges {
                    comment:node {
                      bodyHTML
                    }
                  }
                }
              }
            }
          }
        }
      }
    }

    followers(first: 10) {
      edges {
        follower:node {
          login
        }
      }
    }
  }
}

Calculation:

50              = 50 repositories
 +
50 x 20       = 1,000 pullRequests
 +
50 x 20 x 10 = 10,000 pullRequest comments
 +
50 x 20       = 1,000 issues
 +
50 x 20 x 10 = 10,000 issue comments
 +
10              = 10 followers

                 = 22,060 total nodes

Timeouts

If GitHub takes more than 10 seconds to process an API request, GitHub will terminate the request and you will receive a timeout response and a message reporting that "We couldn't respond to your request in time".

GitHub reserves the right to change the timeout window to protect the speed and reliability of the API.

You can check the status of the GraphQL API at githubstatus.com to determine whether the timeout is due to a problem with the API. You can also try to simplify your request or try your request later. For example, if you are requesting a large number of objects in a single request, you can try requesting fewer objects split over multiple queries.

If a timeout occurs for any of your API requests, additional points will be deducted from your primary rate limit for the next hour to protect the speed and reliability of the API.

Other resource limits

To protect the speed and reliability of the API, GitHub also enforces other resource limitations. If your GraphQL query consumes too many resources, GitHub will terminate the request and return partial results along with an error indicating that resource limits were exceeded.

Examples of queries that may exceed resource limits:

Requesting thousands of objects or deeply nested relationships in a single query.
Using large first or last arguments in multiple connections simultaneously.
Fetching extensive details for each object, such as all comments, reactions, and related issues for every repository.

Query optimization strategies

Limit the number of objects: Use smaller values for first or last arguments and paginate through results.
Reduce query depth: Avoid requesting deeply nested objects unless necessary.
Filter results: Use arguments to filter data and return only what you need.
Split large queries: Break up complex queries into multiple simpler queries.
Request only required fields: Select only the fields you need, rather than requesting all available fields.

By following these strategies, you can reduce the likelihood of hitting resource limits and improve the performance and reliability of your API requests.

Rate limits and query limits for the GraphQL API

本文内容