Add Wikipedia API for failover

Author: plutov

Dockerfile

@@ -4,7 +4,7 @@ FROM python:3
 
 ADD . /
 
-RUN pip install slackclient rasa_nlu scipy scikit-learn sklearn-crfsuite numpy spacy wolframalpha
+RUN pip install slackclient rasa_nlu scipy scikit-learn sklearn-crfsuite numpy spacy wolframalpha wikipedia
 RUN python -m spacy download en
 
 CMD [ "python", "./bot.py" ]

README.md

@@ -1,13 +1,18 @@
 ### Overview
 
-Bot using RASA NLU to answer different questions. Using Wolfram Alpha to get answers.
+Bot using RASA NLU to answer different questions. Using Wolfram Alpha to get answers. Using Wikipedia as failover.
+
+![bot.png](https://raw.githubusercontent.com/plutov/bot/master/bot.png)
 
 ### Run it with Docker
 
 Get Slack API Token first: https://wizeline.slack.com/services/B861V31E3
 
 Get Wolfram App ID: https://developer.wolframalpha.com/portal/myapps/
 
+Wikipedia API doesn't require API key.
+
+
 ```
 docker build -t bot . && docker run -e SLACK_TOKEN=<token> -e WOLFRAM_APP_ID=<app_id> bot
 ```

bot.png

@@ -4,12 +4,12 @@
 import sys
 from slack.bot import SlackBot
 from nlp.rasa import RasaNLP
-from wolfram.wolfram import Wolfram
+from dataprovider.dataprovider import DataProvider
 
 try:
-	w = Wolfram(os.environ.get("WOLFRAM_APP_ID"))
+	dp = DataProvider(os.environ.get("WOLFRAM_APP_ID"))
 
-	r = RasaNLP(w, "rasa-config.json", "rasa-data.json", "./rasa-model")
+	r = RasaNLP(dp, "rasa-config.json", "rasa-data.json", "./rasa-model")
 	r.train()
 
 	b = SlackBot(os.environ.get("SLACK_TOKEN"), r)

bot.py

@@ -0,0 +1,31 @@
+# Copyright (c) 2017 Alex Pliutau
+
+import wolframalpha
+import wikipedia
+import logging
+
+class DataProvider(object):
+	NOT_FOUND_MSG = "Sorry, I don't know this yet"
+
+	def __init__(self, app_id):
+		self.wolfram_client = wolframalpha.Client(app_id)
+
+		logging.basicConfig(level=logging.INFO, format='%(asctime)s %(message)s')
+		logging.info("connected to wolfram")
+
+	def get_short_answer(self, query):
+		logging.info("searching in wolfram: {}".format(query))
+
+		try:
+			wolfram_res = self.wolfram_client.query(query)
+			logging.info("wolfram res: {}".format(wolfram_res))
+
+			return next(wolfram_res.results).text
+		except:
+			# use wikipedia as failover
+			wikiepedia_res = wikipedia.summary(query, sentences=1)
+			logging.info("wikipedia res: {}".format(wikiepedia_res))
+			if wikiepedia_res:
+				return wikiepedia_res
+
+			return self.NOT_FOUND_MSG

wolfram/__init__.py dataprovider/__init__.pywolfram/__init__.py renamed to dataprovider/__init__.py

@@ -19,10 +19,10 @@ class RasaNLP(object):
 	INTENT_WHATIS = "whatis"
 	ENTITY_QUERY = "query"
 
-	def __init__(self, wolfram, config_file, data_file, model_dir):
+	def __init__(self, data_provider, config_file, data_file, model_dir):
 		self.greeted = False
 
-		self.wolfram = wolfram
+		self.data_provider = data_provider
 		self.data_file = data_file
 		self.model_dir = model_dir
 		self.rasa_config = RasaNLUConfig(config_file)
@@ -60,4 +60,4 @@ def find_reply(self, msg):
 		return random.choice(self.COULD_NOT_PARSE_MSGS)
 
 	def get_short_answer(self, query):
-		return self.wolfram.get_short_answer(query)
+		return self.data_provider.get_short_answer(query)