Bot (for Slack et al) with plugins. What's the selling point then? The plugins are Docker containers and the bot interacts with them via stdin/stdout. In plain English that means that you can develop your plugins in whatever language you want and you will just need to pack them as a Docker container to be able to use them with Botella.
Grab a release & give it executable permissions, for example with chmod a+x.
You just need to be sure that you have a botella.yaml (or in another place if you use -f) file in the path where you run it. This file could look like:
adapters:
- name: slack
environment:
key: xoxb-xxx
- name: http
environment:
port: 8080
plugins:
- image: agonzalezro/botella-test
environment:
KEY: this-is-a-secret
only_mentions: trueYou can easily see that we are defining a list of adapter (how to connect with the bot) and plugins that are going to be run when the bot receives a message.
At the moment of writing we support two types of adapters:
To use the slack adapter you just need to define a key in this form:
adapters:
- name: slack
environment:
key: xoxb-xxx
...To get that key you could just go to https://your-org-here.slack.com/services/new/bot and create a new bot, after configuring it you will see an API Token, copy/paste it in your botella.yaml.
Probably, while you were creating the bot in Slack you saw that you could define its profile pic and name, be original!
Note: if for security reasons you prefer to set that API key as an environment variable you can use the environment variable SLACK_KEY.
The HTTP adapter is more easy to setup, you just need to define a port where you want it to be listening:
adapters:
- name: http
environment:
port: 8080 # You can also set HTTP_PORT instead setting it here
...Now you can start POSTing, GETting or whatever to your bot in the / path of wherever you are running it.
For example, if we are running it in localhost in the port 8080 a good way to test it would be using curl:
curl -X POST -d message=hi localhost:8080The plugins is just a list of docker images. Check the previous example:
...
plugins:
- image: agonzalezro/botella-test
environment:
KEY: this-is-a-secret
volumes:
- /tmp:/tmp-from-botella
only_mentions: trueIt has 4 basic sections:
image: it's just the name of the Docker image to be run.environment: the environment variables that you want to set to the container when you run it. Be careful, they are caseSensitive.volumes: the volumes you want to mount from the host that is running botella inside the container that is running the plugin.only_mentionset al. They are basically three permissions that you can set to the plugin and that will work in the adapters that have the concept of channels and direct messages:only_mentions: will make the bot just react when its mentioned.only_channels: will make the bot just react when the message was send in a channel.only_direct_messages: you know how this goes...
Note: If you want you can read the values of the environment keys from the host/system environment keys. Let's explain with an example:
...
plugins:
- image: agonzalezro/botella-test
environment:
KEY:
In the previous configuration we say that we want our container to receive an environment variable called KEY but we don't set any value to it. If we want to give it one value we will just need to run Botella like this:
AGONZALEZRO_BOTELLA_TEST_KEY=xxx ./botella
Of course, how you set those variables is up to you, you don't need to do it inline as explained in the example.
Also note that all the special characters in the image name are being replaced by _s for compatibility reasons.
| Image | Description |
|---|---|
| agonzalezro/botella-test | It's a test plugin that echoes whatever you write to it and shows the value of an environment variable called ENV. |
Please, if you create a cool plugin that can be listed here because it's public, let me know with a PR!
Developing a plugin for Botella is extremely simple, you will need to write a program and pack it with Docker.
That program will receive a JSON and should reply with simple lines. For example, this is a JSON that your program will receive:
{
"version": -1,
"emitter": "U02SLLLH7",
"receiver": "G2TF58RH9",
"body": "u003c@U1PQFQ2SJu003e ping"
}
When your program receives that JSON it will probably check the body to see if it contains the word ping and then return a pong. How do you return a pong? Just write it to the standard output and exit.
In the examples/ folder you can find a simple plugin that does two important things:
- it will echo whatever message is send to it, thinking out of the box you could do whatever you want with that input (from the container stdin) and output whatever you want instead just an echo. It will also use the fantastic jq to parse the JSON.
- it shows the value of an environment variable set on the
botella.yamlfile. What we are trying to show with this? That you can pass secrets around (from thebotella.yamlto the container) without compromise them. Imagine that the container needs an AWS key, just add it to yourbotella.yamland read the value from inside the container.
This plugin is uploaded on my Docker Hub in case you want to test it without building the Docker image: agonzalezro/botella-test.
If you want to debug a plugin that you are working on you can set DEBUG=1 when you run botella and you will see the errors that the plugin container is returning.
Other option is to directly check the logs of that container, you will see the container ID with `docker ps -a .
$ glide install
$ go build$ go test $(glide novendor)The Docker image for botella is based on scratch, the Dockerfile just copies a binary so you need to be sure that the binary is on dist/:
$ ./build-dist.shThen you can build the image:
$ docker build -t [the_name_you_want_to_give] .- Use some interface that avoids calling Docker on the tests.