A Keybase chat bot that notifies a channel when an event happens on a GitLab project (issues, pull requests, commits, etc.).
Prerequisites
In order to run the GitLab bot, you will need
A running MySQL database in order to store user preferences, and channel subscriptions
An arbitrary secret, used to authenticate webhooks from GitLab (this can be any string)
Running
On your SQL instance, create a database for the bot, and run db.sql to set up the tables.
Build the bot using Go 1.13+, like such (in this directory):
go install .
The GitLab bot sets itself up to serve HTTP requests on /gitlabbot plus a prefix indicating what the URLs will look like. The HTTP server runs on port 8080. You can configure nginx or any other reverse proxy software to route to this port and path. Make sure the callback url for your GitLab app is set to http://<your web server>/gitlabbot/oauth.
ngrok provides temporary web urls that can serve from localhost, which means you can use ngrok to test locally. You will need to add your ngrok generated url to the Callback URL section of your GitLab OAuth app. As well as use that as the http-prefix flag when running the bot.
If you accidentally run the bot under your own username and wish to clear the ! commands, run the following:
keybase chat clear-commands
Restricted bots are restricted from knowing channel names. If you would like
a bot to announce or report errors to a specific channel you can use a
ConversationID which can be found by running:
keybase chat conv-info teamname --channel channel
By default, bots are unable to read their own messages. For development, it may be useful to disable this safeguard.
You can do this using --read-self flag when running the bot.
You can optionally save your bot secret inside your bot account's private KBFS folder. To do this, create a credentials.json file in /keybase/private/<YourGitLabBot> (or the equivalent KBFS path on your system) that matches the following format:
{
"webhook_secret": "your secret here"
}
If you have KBFS running, you can now run the bot without providing --secret command line options.
Docker
There are a few complications running a Keybase chat bot, and it is likely easiest to deploy using Docker. See https://hub.docker.com/r/keybaseio/client for our preferred client image to get started.