RASA2.0搭建中文聊天机器人

环境搭建
# 优先安装这几个python包，后面缺什么补充什么
rasa==2.0.0
rasa-sdk==2.4.1
tensorflow==2.3.2
Flask==1.1.2
开始搭建

  • 初始化项目

    # 命令创建rasa项目所需的所有文件，并根据一些示例数据训练一个简单的机器人
rasa init

  • 初始化后会创建以下文件，介绍几个重要的文件，当然这些文件结构你也可以自己调整

    文件名作用
    config.yml配置NLU和Core的文件
    /data/nlu.ymlNLU的训练语句，意图加包含意图的话术
    /data/stories.yml对话的故事，可以理解为剧本
    domain.yml作用域，里面注册你定义的意图，对话模板，自定义action，槽位等
    endpoints.ymlaction的端口，action对外暴露的接口配置文件
    /models/xxxx.tar,gz训练的模型

  • 我项目的目录结构是这样的，可以作为参考

    .
├── actions
│   ├── action_kn_test.py
│   ├── action_trademark_analy.py
│   ├── action_trademark_explain.py
│   ├── action_trademark_keyword.py
│   ├── action_weather_query.py
│   ├── __init__.py
│   └── __pycache__
│       ├── action_kn_test.cpython-36.pyc
│       ├── action_trademark_analy.cpython-36.pyc
│       ├── action_trademark_explain.cpython-36.pyc
│       ├── action_trademark_keyword.cpython-36.pyc
│       ├── action_weather_query.cpython-36.pyc
│       └── __init__.cpython-36.pyc
├── App
│   ├── def_func.py
│   ├── def_settings.py
│   ├── ext.py
│   ├── __init__.py
│   ├── models.py
│   ├── __pycache__
│   │   ├── ext.cpython-36.pyc
│   │   ├── __init__.cpython-36.pyc
│   │   └── views.cpython-36.pyc
│   ├── settings.py
│   ├── static
│   │   ├── css
│   │   │   └── style2.css
│   │   ├── images
│   │   │   ├── ai.png
│   │   └── js
│   │       ├── chat.js
│   │       └── jquery-3.3.1.min.js
│   ├── templates
│   │   └── chat.html
│   └── views.py
├── configs
│   ├── config.yml
│   ├── credentials.yml
│   ├── domain.yml
│   ├── endpoints.yml
│   ├── zh_jieba_mitie_embeddings_config.yml
│   └── zh_jieba_supervised_embeddings_config.yml
├── data
│   ├── dict
│   │   ├── cityname.txt
│   │   └── userdict.txt
│   ├── nlu
│   │   └── nlu.yml
│   ├── rules.yml
│   ├── stories
│   │   ├── stories.yml
│   │   └── story1.md
│   └── total_word_feature_extractor_zh.dat
├── models
│   └── 20210420-153935.tar.gz
├── __pycache__
├── server.py
├── shell_diectory
│   ├── run_shell
│   └── train_shell
开始构建聊天机器人的第一个例子(查天气)

  • 定义你的NLU

    - intent: query_weather
  examples: |
    - 我想查询天气
    - 我想查询[广州]{"entity":"address"}的天气
    - [明天]{"entity":"datetime"}天气怎么样？
    - 我想查询[明天]{"entity":"datetime"}[深圳]{"entity":"address"}的天气

  • 准备地名词库

在这里插入图片描述

  • 定义你stories剧本

    ## query_weather(天气查询)
* query_weather
    - weather_form
    - active_loop{"name":"weather_form"}

  • 在domain.yml 注册，注册你定义的意图，实体，槽位，问话模板，表单，自定义的action等

    version: "2.0"

intents:
  - greet
  - goodbye
  - affirm
  - deny
  - mood_great
  - mood_unhappy
  - bot_challenge
  - query_weather  # 意图查天气


entities:
  - kn_question_num
  - datetime    # 实体时间
  - address     # 实体地址

actions:
  - action_test

slots:
  kn_question_num:
    type: text
  datetime:
    type: text   # 你需要填入的槽位时间
  address:
    type: text	# 你需要填入的槽位地址	
  requested_slot:
    type: unfeaturized

responses:
  utter_greet:
  - text: "Hey! How are you?"

  utter_cheer_up:
  - text: "Here is something to cheer you up:"
    image: "https://i.imgur.com/nGF1K8f.jpg"

  utter_did_that_help:
  - text: "Did that help you?"

  utter_happy:
  - text: "Great, carry on!"

  utter_goodbye:
  - text: "Bye"

  utter_iamabot:
  - text: "I am a bot, powered by Rasa."

  utter_ask_datetime:
  - text: "请问你要查询什么时候的天气？"   # 自动询问模板

  utter_ask_address:
  - text: "请问你要查询哪里的天气？"    #自动询问模板

forms:
  weather_form: {}   #多轮问话表单

session_config:
  session_expiration_time: 60
  carry_over_slots_to_new_session: true

  • 自定义action.py

    ‘’‘
@author: 小帆芽芽
date: 2021/04/25
‘’‘
from rasa_sdk.types import DomainDict
from typing import Dict, Text, Any, List, Union
from flask import jsonify
from rasa_sdk import Tracker, Action
from rasa_sdk.events import UserUtteranceReverted, Restarted, SlotSet, EventType
from rasa_sdk.executor import CollectingDispatcher
from rasa_sdk.forms import FormAction
# from actions import ChatApis
# from actions.WeatherApis import get_weather_by_day
from requests import (
    ConnectionError,
    HTTPError,
    TooManyRedirects,
    Timeout
)
# from test.qachat import get_qa
from rasa_sdk import ActionExecutionRejection
from rasa_sdk.events import SlotSet, AllSlotsReset
from rasa_sdk.executor import CollectingDispatcher
from rasa_sdk.forms import FormAction, REQUESTED_SLOT
import requests, random, math
import re
import json
# import jieba
import logging

logger = logging.getLogger(__name__)

# 查询天气
class QueryWeather(FormAction):
	'''
	心知天气API
	'''
    def __init__(self):
        super().__init__()
        # 天气查询私钥
        self.KEY = '你的私钥'
        self.LOCATION = 'beijing'  # 所查询的位置，可以使用城市拼音、v3 ID、经纬度等
        self.url = 'https://api.seniverse.com/v3/weather/daily.json'  # API URL，可替换为其他 URL
        self.UNIT = 'c'  # 单位
        self.LANGUAGE = 'zh-Hans'  # 查询结果的返回语言
        self.data={
            'key': self.KEY,
            'location': self.LOCATION,
            'language': self.LANGUAGE,
            'unit': self.UNIT,
            'start': 0,
            'days': 10
        }
        # 0 代表今天,1 代表明天,2代表后天
        self.num = 0

    def name(self) -> Text:
        return "weather_form"

    @staticmethod
    def required_slots(tracker):
        return ['address','datetime']
        
	# 槽值映射，不管用户说什么都将填入槽中
    def slot_mappings(self) -> Dict[Text, Union[Dict, List[Dict[Text, Any]]]]:
        return {'address':[self.from_text()],'datetime':[self.from_text()]}
	# 槽值验证
    def validate_address(self,slot_value, dispatcher, tracker, domain):
        # list=['北京','上海','深圳','广州']
        path = './data/dict/cityname.txt'
        d_ = [i.strip() for i in open(path, 'r').readlines()]
        # if slot_value in list:
        if slot_value in d_:
            return {"address":slot_value}
        else:
            dispatcher.utter_message('您输入的不是地名哟')
            return {"address":None}
    # 槽值验证
    def validate_datetime(self,slot_value, dispatcher, tracker, domain):
        d_list = ['今天','明天','后天']

        if slot_value in d_list:
        # if slot_value in d_:
            return {"datetime":slot_value}
        else:
            dispatcher.utter_message('暂不支持查询 {} 的天气'.format(slot_value))
            return {"datetime":None}

    def submit(self,dispatcher, tracker, domain):
        date = tracker.get_slot('datetime')
        address = tracker.get_slot('address')
        self.judge_time(date)
        text_day, text_night, high, low, wind_direction, humidity, wind_speed, wind_scale,date_ = self.get_weather(address)
        s='{}{}({})的天气情况是：白天{},晚上{}.当天最高温度是{}摄氏度,最低温度是{}摄氏度.风向{},风速{},风力等级为{},空气湿度为{}%'.format(
           address,date,date_, text_day,text_night,high,low,wind_direction,wind_speed,wind_scale,humidity
        )
        dispatcher.utter_message('小娜为您查询到：')
        dispatcher.utter_message(s)

        # return [AllSlotsReset()]
        return [Restarted()]


    def judge_time(self,date):
        if date == '今天':
            self.num = 0
        elif date == '明天':
            self.num = 1
        elif date == '后天':
            self.num = 2

    def get_weather(self,address):
        self.LOCATION = address
        str_ = requests.get(url=self.url,params=self.data,timeout=2).text
        dic = json.loads(str_)
        print(dic)
        # 白天天气
        text_day = dic['results'][0]['daily'][self.num]['text_day']
        # 夜晚天气
        text_night = dic['results'][0]['daily'][self.num]['text_night']
        # 最高温度
        high = dic['results'][0]['daily'][self.num]['high']
        # 最低温度
        low = dic['results'][0]['daily'][self.num]['low']
        # 风向
        wind_direction = dic['results'][0]['daily'][self.num]['wind_direction']
        # 湿度（百分比）
        humidity = dic['results'][0]['daily'][self.num]['humidity']
        # 风速
        wind_speed = dic['results'][0]['daily'][self.num]['wind_speed']
        # 风力等级
        wind_scale = dic['results'][0]['daily'][self.num]['wind_scale']
        # 日期
        date_ = dic['results'][0]['daily'][self.num]['date']

        return text_day,text_night,high,low,wind_direction,humidity,wind_speed,wind_scale,date_

  • 采用flask外部调用接口

    '''
@author:小帆芽芽
@date:2021/04/25
'''
import json
import re

import requests
from flask import Blueprint, render_template, request

blue = Blueprint('rasa', __name__)

REQUEST_URL = "http://localhost:5005/webhooks/rest/webhook"
HEADERS = {'Content-Type': 'application/json; charset=utf-8'}


def init_route(app):
    app.register_blueprint(blue)

@blue.route('/', methods=['GET', 'POST'])
def test():
    # return 'this a test page', 200
    return render_template('chat.html')


@blue.route('/chat',methods=['GET'])
def chat():
    data=request.args.to_dict()
    message=data['message']
    if message:
        answer = rasaresponse('xiaofangyaya',message)
        return answer

def rasaresponse(user,msg):
    requestDict={'sender':user,'message':msg}
    rsp=requests.post(REQUEST_URL, data=json.dumps(requestDict),headers=HEADERS)
    # content=rsp.text.encode('utf-8').decode("unicode-escape")
    if rsp.status_code == 200:
        rsp_json=json.loads(rsp.text.encode())
        content_text = ''
        if len(rsp_json):
            # print(rsp_json)
            for i in range(len(rsp_json)):
                # txt_=rsp_json[i]['text']
                txt_=rsp_json[i].get('text')
                print('rasa返回内容是---------->{}'.format(txt_))
                if txt_:
                    content_text=content_text+txt_
                else:
                    content_text=content_text + '\n' + rsp_json[i]['image']

            return content_text

        else:
            return '哎呀,这个小娜也不懂呢～～～'

    else:
        return '返回错误，请重新输入问话。'

  • 运行rasa，依次执行以下命令

    # 启动服务
# 1. 启动rasa server
python -m rasa run --port 5005 -m models/20210420-153935.tar.gz --endpoints configs/endpoints.yml --credentials configs/credentials.yml --debug
# 2. 启动action server
python -m rasa run actions --port 5055 --actions actions --debug
# 3. 启动server.py
python server.py

  • 页面测试

在这里插入图片描述

在这里插入图片描述

  • 我的源码地址：
    https://gitee.com/menfeng/rasa-weather-edition

希望点赞关注支持哦！！！！
后期还有大波福利发送
在这里插入图片描述

# python # flask # 人工智能 # tensorflow
Logo
AtomGit 开源协作平台测评赛

瓜分20万奖金 获得内推名额 丰厚实物奖励 易参与易上手

更多推荐

cover

C#联合Halcon深度学习源代码分享1 预处理图像2图像识别测试3误差分析（含导入步骤文档，含中文注释） （附源码链接）

avatar 开放原子开发者工作坊

Git基础命令学习

git基础命令学习笔记git init 命令目录变成 Git 可以管理的仓库git add 把文件添加到仓库(可多次add不同的文件)git commit 把文件提交到仓库git satus 命令查看状态，可以让我们时刻掌握仓库当前的状态git diff可以看到指定文件的修改内容

avatar 开放原子开发者工作坊
cover

SpringBoot01:Hello,World

avatar 开放原子开发者工作坊