widenbot
A Slack bot and plugin framework for Widen
| |
% docker build -t widen/widenbot .
% docker run -it --rm --name widenbot widen/widenbot
(IMHO, this is the easiest way to get started, and mirrors the way in works
in production).
Global Install:
% npm install -g https://github.com/widen/widenbot
% cp config.example.js config.js
% widenbot -c config.js
Local Install (recommended):
% npm install https://github.com/widen/widenbot
% cp config.example.js config.js
% ./node_modules/.bin/widenbot -c config.js
It is recommended to use environment variables for secret data such as API
keys and secrets.
- Plugin framework: easily create new plugins to do anything in response to
Slack commands!
- Webhook framework: send messages to users or channels when another service
alerts the bot.
- Logging: Extensible logging framework. Can send data to loggly, or anywhere
really.
The configuration for the bot is stored in a javascript file. That is
require
d by the bot process. This means you can include executable Javscript
within the configuration. An example is provided in config.example.js
The configuration is loaded and interpreted at runtime. A configuration can be
provided via the command-line wrapper.
% widenbot --config path/to/config.js
Currently recognized fields are:
name
: the bot's nameicon
: the bot's icon. Can be an emoji (":ghost:") or a URL.repl
: whether to launch an interactive repl with all plugins loaded
(default: false)token
: Slack's outgoing webook tokenurl
: Slack's incoming webhook urldbpath
: Path for the database to exist atlogging
: Options for the logger
path
: Path for the logs to be written toconsole
(optional): If true logs are written to console too
plugins
: A hash of plugin names and options. Only plugins in this hash
will be recognized.
The configuration file should simply export a plain ol' Javascript object with
keys and values corresponding to the desired configuration values.
module.exports = {
"key": "value",
"key2": ["array_value1", "array_value2"],
"env_key": process.env.SECRET_KEY_NAME
};
The brain is the memory for the bot. This is persisted using levelup and
leveldown (leveldb) with an eye for supporting any sort of backend in the
future.
Each plugin is given a namespace within the brain to store and retrieve data.
Required options:
dbpath
: Path for the database to exist at
Plugins can extend the bot's ability to respond and act on commands.
Required options:
plugins
: A hash of plugin names and options hashes. Only plugins in this hash
will be recognized.
Example:
plugins: {
"echo": {}
"lastfm": {
"api_key": "BLAHBLAH"
}
}
The plugin name ("echo" or "lastfm" in the example) must correspond to the filename in ./plugins
, or a node module that can be require
'd'.
The plugin options must be either the empty object {}
, or a hash of
options. These options will be avaiable to the plugin when it is evaluated so
integrators can access API keys and other configuration variables for that
plugin.
widenbot uses bunyan as its logging backend since it integrates nicely
with restify (the web framework being used).
Options for logging are specified in the logging
hash in the configuration.
logging
: Options for the logger
path
: Path for the logs to be written toconsole
(optional): If true logs are written to console too
Plugins are easy to create. Just add a new file to ./plugins
or install a
compatible node module, and add the plugin name and options to the
configuration.
A plugin must export a plain ol' Javascript object with a few required
properties:
pattern
:
this is the regex pattern that incoming commands will be matched against.
If a match is found, then this plugin will execute on that command.
respond
:
This is the main response handler function for the plugin. It
receives a context
(which is an object with the command, username,
arguments, plugin options, and a reference to the brain).
This function should return a value or a promise. If the value is empty string, then
nothing is sent back. If promise is rejected, then nothing is sent back.
Else, resolve value is sent back, or the returned value is sent back.
The response can also add attachments to the ctx.outgoing_message
as outlined in the Slack API.
# Must export an object '{}'
var Echo = module.exports = {
# metadata
"name": "Echo", # Plugin's name
"author": "Mark Feltner", # Plugin author's name
"description": "Echoes what was just said.", # Plugin description
"help": "echo", # Plugin help text
# what text the command should match
"pattern": /^echo$/,
# function that will respond
# can return a value or a promise
"respond": function(ctx) {
return ctx.args;
}
};
Much like plugins, webhooks are POJOs defined in the configuration.
Rather than match a regex like plugins, webhooks are matched via their name
and a url namespace. All webhook requests should follow the format:
http://path-to-bot/webhooks/:webhook_name
where webhook_name is the name
of the hook defined in config.js
, and all webhook requests should be POSTs.
respond
:
Responds to the webhook. Should follow the signature:
function (context, req, res, next){}
As of right now, must return a promise. The resolve()
success value
is hash containing:
{
to: '',
response: '',
attachments
from: '',
in the config
}
% git clone git@github.com:Widen/widenbot.git
% cd widenbot
% npm i
% cp config.example.js config.js
% vim config.js
% node bin/cli.js -c config.js
All development build scripts are located in the package.json
scripts
key.
npm run changelog
: (WIP) generate a changelognpm run watch
: if you have nodemon this will reload on changesnpm run lint
: Run jshint on codenpm run unit-test
: Run test suitenpm run coverage
: Run coverage suietnpm test
: Run lint and test suite (this is used by Travis CI)
See our contributing guidelines