class SearchHubLogicAdapter(LogicAdapter):
def __init__(self, **kwargs):
super(SearchHubLogicAdapter, self).__init__(**kwargs)
self.host = kwargs.get("searchhub_host")
self.port = kwargs.get("searchhub_port")
self.shub_url = "http://{0}:{1}/api/apollo/query-pipelines/lucidfind-default/collections/lucidfind/select".format(self.host, self.port)
self.shub_display_url = "http://{0}:{1}/p:%20?{2}" #note the %20 is a hack around how shub handles the projects, eventually we'll pass in filters here
print "URL: {0}".format(self.shub_url)
self.closest = ClosestMeaningAdapter()
def can_process(self, statement):
#print statement.extra_data
is_q = False
is_bot_ask = False
if "is_question" in statement.extra_data:
is_q = statement.extra_data["is_question"]
if "is_bot_ask" in statement.extra_data:
is_bot_ask = statement.extra_data["is_bot_ask"]
if statement.text.lower().startswith("shub:"):
return True
return is_q or is_bot_ask
def process(self, statement):
print "calculating: {0}, {1}".format(statement, statement.extra_data)
result = None
input_channel = statement.extra_data["slack_channel"]
is_q = statement.extra_data["is_question"]
is_bot_ask = statement.extra_data["is_bot_ask"]
text = statement.text
if statement.text.lower().startswith("shub:"):#If the user specifically asks for 'searchhub' (shub), then strip it off
text = text[5:]
is_bot_ask = True
params = {
"wt":"json",
"q": text,
"rows": 3,
"fl": "id,title,subject,body,score",
"fq":"isBot:false"
}
#TODO: filter projects based on channels
print "Params: {0}".format(params)
response = requests.get(self.shub_url, params)
if response.status_code != 200:
return 0, None
rsp = response.json()
#let's see if we have a decent response
confidence = 0
if "response" in rsp:
#print rsp
num_found = rsp["response"]["numFound"]
print num_found
if num_found > 0:
url = self.shub_display_url.format(self.host, self.port, urllib.urlencode({"q": text}))
response = "SearchHub says...\n\tSee full results: {0}\n\n".format(url)
docs = rsp["response"]["docs"]
#print "docs: {0}".format(docs[0])
#confidence = docs[0]["score"]
#Confidence is the average of the similarity scores, for now
i = 0
for doc in docs:
display = ""
if "title" in doc:
display = doc["title"]
elif "subject" in doc:
display = doc["subject"]
#see how similar the display value is to the original
similarity = self.closest.get_similarity(text, display)
print similarity
confidence += similarity#should we add here? Probably not, but let's try it
response += "{0}:\n\t{1}\n\t{2}\n".format(i, doc["id"], display.encode('utf-8'))
i += 1
result = Statement(response)
#pass through the metadata, so we have channel info
if i > 0:
confidence = confidence / i
if is_bot_ask: # boost confidence if the user specifically invoked the kraken
confidence += 10
else:
print "Couldn't find results for {0}".format(params)
result = Statement("Shub no find answer.")
confidence = 0
#print "Conf: {0}, Res: {1}".format(confidence, result)
else:
#print "Couldn't find results for {0}".format(params)
result = Statement("Shub no find answer.")
confidence = 0
result.extra_data = statement.extra_data
return confidence, result
