Skip to main content

Comments > Author type

Written by Tom Williams

Dataset: Comments

Entity: Comments

Field ID: author_type

Type: Select list

Description: The type of the author of the comment. Possible values are:

  • PERSON a person.

  • BOT a bot or an app: a known review, CI or automation bot (e.g. CodeRabbit, Dependabot, GitHub Actions), or an account the application itself flags as a bot.

  • TEAM a team or group, such as a team requested as reviewer. Rare as an author.

Leave bots out with one filter. author_type != "BOT" keeps the comments written by people and teams, without naming each bot. A comment whose author is unknown has an empty author type, and this filter leaves it out too.

Source: Calculated

Transformation logic: Keypup types every account from two sources, and an account is a BOT when either says so:

  • The application's own account type, where the application gives one: see the App Mapping below. Otherwise the account is a PERSON.

  • A list of known bots maintained by Keypup, kept per application so that a name that is a bot on one application can be a person on another. It covers AI code reviewers and coding agents, code quality and security scanners, dependency updaters, CI and deployment services, and the automation apps of work trackers: CodeRabbit and Sonarcloud on Azure DevOps, or Automation for Jira, for example. On every application, an account whose name ends in [bot], -bot, _bot or .bot is a bot as well.

Comments copied from another tool are bot comments. A sync app (e.g. Exalate or Unito on Jira) posts the comments it copies under its own account, at the time of the sync rather than of the comment. Jira flags that account as an app, so these comments are typed BOT: they credit no one on your team.

A bot of your own counts as a person unless its name says otherwise. The list holds products used across many teams, not a team's own release, CI or sync accounts. Name such an account with a bot suffix, e.g. release-bot or PR [bot], and it is typed BOT. If a product bot is shown as a PERSON, contact support and we will add it to the list for everyone.

App Mapping

GitHub (Comments)

The account type GitHub gives: a user is PERSON, an app (such as dependabot or github-actions) is BOT, a team is TEAM.

GitLab (Comments)

The account's bot flag: project and group access token users and service accounts are BOT.

Bitbucket (Comments)

The account type: app users (Cloud) and service accounts (Data Center) are BOT, team accounts are TEAM.

Azure DevOps (Comments)

The identity type: service principals and service identities (such as the build service) are BOT, groups are TEAM. Review and analysis tools that sign in as ordinary accounts, such as CodeRabbit and Sonarcloud, are BOT through Keypup's list.

Jira (Comments)

The account type: apps (such as Automation for Jira, or a sync app) are BOT, Atlassian accounts and service desk customers are PERSON.

ClickUp (Comments)

Calculated: Keypup's list only, ClickUp gives no account type.

Trello (Comments)

Calculated: Keypup's list only, Trello gives no account type.

Reporting Use Cases

The author_type field separates the work of people from the work of automation. As a select list, it is mostly used as a filter, and sometimes as a dimension.

  • Peer comments without bots: a peer comment is one written by someone other than the author of the issue or pull request. COUNT_IF(author_type != "BOT" && author_username != parent_author_username) / IF_ZERO(COUNT_DISTINCT(parent_system_id), 1) gives the number of peer comments per item, without the AI reviewers and scanners that can write dozens of comments on a single pull request.

  • How much of the conversation is automated: a bar chart with author_type as the dimension and COUNT() as the metric, filtered on parent_type == "pull_request", shows how pull request discussion splits between people and bots.

  • Which bots comment the most: filter on author_type == "BOT" and group by author_username.

  • The author of the parent review: review_author_type (Review > Author type) holds the type of the author of the review a comment belongs to, with the same values.

  • Time to first human feedback on an item: the Issues & Pull Requests dataset carries first_peer_comment_at, the first comment by someone who is neither the author nor a bot, on the same row as the item's other dates.

Did this answer your question?