大家所熟知的智能聊天机器人,比如:微软小冰、Apple Siri、Google Now、IBM Watson等。微信自动回复、微信服务号里的客服Bot也都算是简单的实例。聊天机器人(Dialog System)主要实现方式有:基于人工模板(比如AIML)、基于搜索技术(比如ElasticSearch)、基于深度学习。从零实现这样的系统是很复杂的。
开源的ChatOps机器人:
1.Hubot:CoffeeScipt编写
2.Lita:Ruby编写
3.Errbot:Python编写
So, What is ChatOps? And How do I Get Started?
12+ Frameworks to Build ChatOps Bots
Hubot是由Github开发的开源聊天机器人,基于Node.js采用CoffeeScript编写。
Hubot https://hubot.github.com/
Hubot Scripts https://github.com/hubot-scripts
Hubot Control https://github.com/spajus/hubot-control
可以借助Hubot开发Chatbot来自动化的完成想要一切自动化任务,比如:
-运维自动化(编译部署代码、重启机器,监控服务器运行情况,自动修复Bug等)
-外部服务交互(管理Redmine、集成Jenkins、监视Zabbix等)
-定时获取天气预报
-随机订餐
-聊天机器人等等。
官方文档里有详细的使用说明,https://hubot.github.com/docs/,这里只做一个摘要。
Hubot Scripts里有大量的现成脚本可以用,也是自己编写脚本的最好sample。
(一)安装
运行Hubot需要以下软件支持:
- Node.js
- Redis 默认存贮数据
- CoffeeScript
- Yeoman
- generator-hubot 生成Hubot骨架工程
v0.12.8
C:\Users\rensanning>npm -v
2.14.9
C:\Users\rensanning>npm install -g yo generator-hubot
C:\Users\rensanning>npm list -g generator-hubot yo
(二)创建自己的bot
D:\>mkdir hubotsample
D:\>cd hubotsample
D:\hubotsample>yo hubot
? Owner RenSanNing <rensanning@gmail.com>
? Bot name okbot
? Description A sample hubot
? Bot adapter campfire
默认提供了campfire和shell的adapter,其他的需要通过npm下载hubot-<AdapterName>。
也可以通过参数直接生成:
项目结构:
- bin/ 运行脚本
- node_modules 应用的包文件
- scripts 存放自定义脚本
- external-scripts.json 应用的外部脚本
- hubot-scripts.json *** 该文件已经无用,可以删除
- package.json 项目全局配置信息
(三)运行bot
2.19.0
D:\hubotsample>bin\hubot -h
Usage hubot [options]
Available options:
-a, –adapter ADAPTER The Adapter to use
-c, –create PATH Create a deployable hubot
-d, –disable-httpd Disable the HTTP server
-h, –help Display the help information
-l, –alias ALIAS Enable replacing the robot’s name with alias
-n, –name NAME The name of the robot in chat
-r, –require PATH Alternative scripts path
-t, –config-check Test hubot’s config to make sure it won’t fail at startup
-v, –version Displays the version of hubot installed
-a 指定Adapter(默认是shell)
-d 关闭HTTP服务(默认是开启的)
-c deprecated 使用yeoman
采用默认的shell adapter
指定adapter
usage:
history
exit, \q – close shell and exit
help, \? – print this usage
clear, \c – clear the terminal screen
okbot> Shell: okbot adapter – Reply with the adapter
okbot animate me <query> – The same thing as `image me`, except adds a few parameters to try to return an animated GIF instead.
okbot echo <text> – Reply back with <text>
okbot help – Displays all of the help commands that Hubot knows about.
okbot help <query> – Displays all help commands that match <query>.
okbot image me <query> – The Original. Queries Google Images for <query> and returns a random top result.
okbot map me <query> – Returns a map view of the area returned by `query`.
okbot mustache me <url|query> – Adds a mustache to the specified URL or query result.
okbot ping – Reply with pong
okbot pug bomb N – get N pugs
okbot pug me – Receive a pug
okbot the rules – Make sure hubot still knows the rules.
okbot time – Reply with current time
okbot translate me <phrase> – Searches for a translation for the <phrase> and then prints that bad boy out.
okbot translate me from <source> into <target> <phrase> – Translates <phrase> from <source> into <target>. Both <source> and <target> are optional
ship it – Display a motivation squirrel
okbot ping
okbot> PONG
okbot echo 你好!
okbot> 你好!
okbot time
okbot> Server time is: Fri Sep 30 2016 11:05:24 GMT+0800 (中国 (标准时间))
***所有的输入都会被记录在.hubot_history文件里。
(四)编写脚本
编写的自定义脚本要放在scripts中,可以是.coffee或.js文件。
scripts/hello.coffee
- # Description:
- # This is a test.
- #
- # Commands:
- # okbot helo – Reply with world!
- module.exports = (robot) ->
- robot.respond /hello/i, (msg) ->
- msg.send “world!”
okbot> world!
因为Hubot要解析脚本文件,提供help帮助,所以脚本文件开头的注释是规范的,
第一行必须是注释“# Description:”(其他的可以没有),否则会有警告:
Hubot同时也支持js,比如:
scripts/hello2.js
- // Description:
- // This is a test2.
- // Commands:
- // okbot helo – Reply with world!
- module.exports = function(robot) {
- robot.respond(/hi/i, function(msg){
- msg.send(“world2!”);
- });
- }
okbot> world2!
Respond vs Hear
- respond只监听直接发送给机器人的消息,需要指定机器人名称
- hear可以监听任何消息
myhubot xxx
@myhubot xxx
myhubot: xxx
Send vs Reply
- send会将消息发送给所有人
- reply会将消息回复给指定的人
Random
msg对象有一个random方法,可以从之后的数组对象中随机提取一个元素
hubot Scripts Explained
http://theprogrammingbutler.com/blog/archives/2011/10/28/hubot-scripts-explained/
(五)安装脚本
Hubot 有一大堆现成的脚本,可以集成各种服务。
D:\hubotsample>npm install –save hubot-plusplus
hubot-plusplus@1.3.0 node_modules\hubot-plusplus
├── underscore@1.8.3
├── clark@0.0.6
└── coffee-script@1.6.3
将package-name添加到external-scripts.json
okbot> ruby has 1 point
okbot> java–
okbot> java has -1 points
(六)hubot-script实例
定时脚本
scripts/cron.coffee
- cronJob = require(‘cron’).CronJob
- module.exports = (robot) ->
- send = (room, msg) ->
- response = new robot.Response(robot, {user : {id : -1, name : room}, text : “none”, done : false}, [])
- response.send msg
- new cronJob(‘0 * * * * *’, () ->
- currentTime = new Date
- send ‘#your-channel-name’, “current time is #{currentTime.getHours()}:#{currentTime.getMinutes()}.”
- ).start()
D:\hubotsample>bin\hubot -a shell
http请求
scripts/googleGEO.coffee
- module.exports = (robot) ->
- robot.hear /location (.*)/, (msg) ->
- request = robot.http(“https://maps.googleapis.com/maps/api/geocode/json”)
- .query(address: msg.match[1])
- .get()
- request (err, res, body) ->
- json = JSON.parse body
- location = json[‘results’][0][‘geometry’][‘location’]
- msg.send “#{location[‘lat’]}, #{location[‘lng’]}”
okbot> 39.904211, 116.407395
抓取数据(request, cheerio)
scripts/title.coffee
- request = require ‘request’
- cheerio = require ‘cheerio’
- module.exports = (robot) ->
- robot.respond /title (.*)/i, (msg) ->
- url = msg.match[1]
- options =
- url: url
- timeout: 2000
- headers: {‘user-agent’: ‘node title fetcher’}
- request options, (error, response, body) ->
- $ = cheerio.load body
- title = $(‘title’).text().replace(/\n/g, ”)
- msg.send(title)
D:\hubotsample>npm install –save cheerio
D:\hubotsample>bin\hubot -a shell
okbot> okbot title http://github.com
okbot> How people build software · GitHub
okbot> okbot title http://www.google.com
okbot> Google
http应答(httpd)
scripts/version.coffee
- module.exports = (robot) ->
- robot.router.get “/version”, (req, res) ->
- res.end robot.version
访问http://localhost:8080/version。
默认端口是8080,可以修改环境变量:export PORT=8080
Hubot大量依赖环境变量来配置脚本,所以一般都做一个启动脚本:
export HUBOT_ENV_TEST_VAR=””
bin/hubot -a twitter -n testbot
SET HUBOT_ENV_TEST_VAR=””
bin\hubot.cmd -a twitter -n testbot
脚本中的使用:
捕获所有未处理信息
scripts/catchAll.coffee
- module.exports = (robot) ->
- robot.catchAll (res) ->
- res.send “Nothing Found:#{res.message.text}”
hubotスクリプトの書き方とサンプル集
http://blog.fumiz.me/2012/08/05/hubot-irc-bot-script/
编写 Hubot Scripts
http://scarletsky.github.io/2016/05/02/write-your-own-hubot-scripts/
(七)自定义Adapter
hubot默认提供两种adapter:shell、campfile
shell用于开发调试,campfile以外的Chat Service也都有开源的实现。
Adapter基本构成
新建文件 \node_modules\hubot\src\adapters\SampleAdapter.coffee
- class SampleAdapter extends Adapter
- send: (envelope, strings…) ->
- @robot.logger.info “Send”
- run: ->
- @robot.logger.info “Run”
- exports.use = (robot) ->
- new SampleAdapter robot
修改文件 \node_modules\hubot\src\robot.coffee
‘campfire’
‘SampleAdapter’
‘shell’
]
启动bot
* 所有的Adapter必须继承自Adapter
* Adapter中最重要的两个方法是send和run,其他方法比如emote、reply、topic、play等在特定场景下才需要。
* run方法是bot启动时执行
* send方法是回复信息时执行
* 父类 adapter.coffee 里有详细的方法说明 https://github.com/github/hubot/blob/master/src/adapter.coffee
* Adapter名必须是:src\adapters里的文件名 & robot.coffee的HUBOT_DEFAULT_ADAPTERS里定义的名 或者和 hubot-#{adapter} 一致
Chat服务的Adapter
- {Adapter, TextMessage} = require ‘hubot’
- {EventEmitter} = require ‘events’
- class MyChatAdapter extends Adapter
- send: (envelope, strings…) ->
- @bot.send str for str in strings
- run: ->
- options =
- token: process.env.HUBOT_CHAT_TOKEN
- rooms: process.env.HUBOT_CHAT_ROOMS
- account: process.env.HUBOT_CHAT_ACCOUNT
- bot = new MyChatStreaming options, @robot
- bot.on ‘message’, (userId, userData, message) ->
- user = @robot.brain.userForId userId, userData
- @receive new TextMessage user, message
- bot.listen()
- exports.use = (robot) ->
- new MyChatAdapter robot
- class MyChatStreaming extends EventEmitter
- constructor: (options, @robot) ->
- @token = options.token
- @rooms = options.rooms.split(“,”)
- @account = options.account
- send: (message) ->
- # Send data to your chat service
- listen: ->
- # Get messge data from chat service
- # @emit ‘message’, user, message
具体可以参考: https://github.com/github/hubot/blob/master/src/adapters/campfire.coffee
扩展robot的方法:
\node_modules\hubot\src\adapters\incircle.coffee
- class TestXXX extends Adapter
- constructor: (robot) ->
- super robot
- robot.hearXXX = (options, callback) ->
- robot.listeners.push new TextListener(robot, “@XXXmessage”, options, callback)
\scripts\testbot.coffee
- module.exports = (robot) ->
- robot.hearXXX (msg) ->
- msg.send “#{JSON.stringify msg}”
(八)微信adapter
https://github.com/KasperDeng/Hubot-WeChat
主要机制是hack网页版微信协议,先用手机登录微信帐号,然后模拟网页版微信登录,这样就可以接受微信消息了。
- npm install hubot-weixin –save
- 手机端登录微信
- 打开网页微信:web.weixin.qq.com
- 手机扫描登录
- 控制台查看如下信息 /node_modules/hubot-weixin/config.yaml
- bin\hubot.cmd -n bot -l / -a weixin
cookie: Uin: Sid: Skey: DeviceID:
参考:
正在吃掉世界的Bot:它从哪里来,会到哪里去?
Hubot:来自GitHub的聊天机器人
GitHub社謹製! bot開発・実行フレームワーク「Hubot」
TDD Hubot scripts with gulp+mocha
在Skype中使用Hubot
http://qiita.com/bouzuya/items/c7d0ad80c357aab6b696