StockBot

StockBot is a chatbot based on Rasa NLU aims to provide stock and weather information implemented by Python.

 

Demo Video

If the video is unable to load, you can download it: Demo.mp4

 

Environment

• Python 3.6.7
• Jupyter Notebook

 

Structure

└── StockBot                        
    ├── StockBot.ipynb         // chatbot
    ├── training_data.json     // training data for extracting intents and entities
    ├── config_spacy.yml       // config for trainer
    ├── city_code.db           // database(all provinces in China with their codes)
    ├── Report.pdf             // report of this project
    └── Demo.mp4               // a demo showing an example dialogue

 

Requirements

pip install rasa_nlu
pip install iexfinance
pip install wxpy
pip install sqlite3
pip install matplotlib

 

How to run

Just download the project and run it in jupyter notebook.

 

Usage

Rasa NLU

Documentation: https://rasa.com/docs/nlu/

Import:

from rasa_nlu.training_data import load_data
from rasa_nlu.config import RasaNLUModelConfig
from rasa_nlu.model import Trainer
from rasa_nlu import config

Train

trainer = Trainer(config.load("config_spacy.yml"))
training_data = load_data('training_data.json')
interpreter = trainer.train(training_data)

Extract intents and entities

interpreter.parse("message")

 

Iexfinance

Documentation: https://addisonlynch.github.io/iexfinance/devel/

Import:

from iexfinance.stocks import Stock
from iexfinance.stocks import get_historical_data
from iexfinance.stocks import get_historical_intraday

Get data:

Stock(company).get_price() # get current price of a certain stock
Stock(company).get_key_stats()['ttmEPS'] # get TTM of a certain stock
Stock(company).get_news() # get relevant news of a certain stock

 

Weather API

First you need to apply for a key in: https://www.juhe.cn/docs/api/id/39 (Documentation can also be found there), and put it here:

def get_weather(day_list, city):
    weather_key = ""
    
    # actions...

This API needs you to post a request with the code of the city or province which you want to query about. In order to try some database operations, this project establish a database consisting of all provinces in China with their codes (city_code.db).

def get_citycode(city):
    conn = sqlite3.connect('city_code.db')
    c = conn.cursor()
    query = "SELECT * FROM city WHERE name = '" + city + "'"
    c.execute(query)
    result =  c.fetchall()
    
    code = '' 
    for row in result:
       code = row[0]
    return code

In fact this database is not necessary. If you don't want to do these database operations, you can try this API providing each province and city’s code: https://www.juhe.cn/docs/api/id/39.

 

Connect to Wechat

wxpy

This part is based on the Python package: wxpy. Documentation: https://wxpy.readthedocs.io/zh/latest/bot.html.

from wxpy import *

bot = Bot()
my_friend = bot.friends().search("name_of_your_friend")[0]

@bot.register(my_friend, TEXT)
def auto_reply(msg):
    
    # actions...

bot.registered

A QR code will be generated after creating a bot. Use your WeChat account to scan this QR code and login, then you can use your friend's WeChat accout to chat with this chatbot.

itchat

Because wxpy has not been updated for a long time, this project provides another way to connect to WeChat through a Python package called itchat. Documentation: https://itchat.readthedocs.io/zh/latest/

import itchat
from itchat.content import *

itchat.auto_login()
my_friend = itchat.search_friends(name="name_of_yourfriend")[0]['UserName']

@itchat.msg_register([TEXT])
def auto_reply(msg):
    
    # action...

itchat.run()

 

Debug in terminal

You can debug the code in terminal in following way instead of connecting to WeChat:

while True:
    state = MAIN
    pending = None

    msg = input()
    print("USER: " + msg)

    state, pending, final_response, message_intent = send_message(state, pending, msg)
    print("BOT: " + final_response)

Have fun :)