Abbot - qq for Slack by Alumstabase

an Abbot package for Slack by Alumstabase

QQ is a question & answer bot.

Abbot is a programmable bot that turns your team chat into a shared command center. We handle all the boilerplate of building and running these conmmands so that you can focus on making tools that help you ship faster.

We built Abbot because we saw the power of this style of work (called ChatOps), when we worked at GitHub. ChatOps made it possible for GitHub to work productively without meetings, while globally distributed. We think it’s a pretty great way to work, so we made it easy to use in Slack, Discord, and Microsoft Teams.

You can read more about Abbot here, check out our blog, or take a look at some of the other cool packages available as a one-click install from Abbot’s Package Directory.

# Change the code below with the code for your skill! import re import requests from typing import Tuple, List class MODELS: QA = “st_msmarco_distilbert_base_v3” SIMILARITY = “st_paraphrase_distilroberta_base_v1” class NLUDB: pass class NLUDBEmbeddingIndex: “”” An Embedding Index is a persistent index over embeddings. “”” def __init__(self, nludb: NLUDB, id: str, name: str): self.nludb = nludb self.name = name self.id = id def insert(self, value: str): return self.nludb.post(’embedding-index/insert’, { “indexId”: self.id, “value”: value }) def embed(self): return self.nludb.post(’embedding-index/embed’, { “indexId”: self.id }) def search(self, query: str): return self.nludb.post(’embedding-index/search’, { “indexId”: self.id, “query”: query }) class NLUDB: def __init__(self, api_key: str=None, version: int=1): self.api_key = api_key self.endpoint = “https://api.nludb.com/api/v{}”.format(version) def post(self, operation: str, payload): url = “{}/{}”.format(self.endpoint, operation) resp = requests.post( url, json=payload, headers = {“Authorization”: “Bearer {}”.format(self.api_key)}) j = resp.json() if ‘data’ in j: return j[‘data’] else: raise Exception(j[‘reason’]) def create_index(self, name: str, model: str, upsert: bool = False): res = self.post(’embedding-index/create’, { “name”: name, “model”: model, “upsert”: upsert }) return NLUDBEmbeddingIndex(nludb=self, name=name, id=res[“id”]) class NludbChatbotBase: def __init__(self, api_key: str, server_name: str, min_confidence: float = 0.4): self.nludb = NLUDB(api_key) self.server_name = server_name self.min_confidence = min_confidence self.index = self.nludb.create_index( server_name, MODELS.QA, upsert=True ) def get_fact(self, message: str) -> str: words = re.sub(“\s\s+” , ” “, message).split(‘ ‘) if len(words) == 0: return None first = words[0].lower() if first in [‘fact’, ‘fact:’, ‘learn’, ‘learn:’]: return ” “.join(words[1:]) return None def dispatch(self, message: str, username: str = None, roomname: str = None) -> str: “”” Todo: add a “forget” feature “”” maybe_fact = self.get_fact(message) if maybe_fact is not None: return self.add_fact(maybe_fact, username, roomname) return self.answer_question(message, username, roomname) def add_fact(self, fact: str, username: str = None, roomname: str = None) -> str: self.index.insert(fact) self.index.embed() return None def answer_question(self, question: str, username: str = None, roomname: str = None) -> Tuple[float, str]: resp = self.index.search(question) if ‘hits’ not in resp or len(resp[‘hits’]) == 0: return None confidence = resp[‘hits’][0][‘score’] text = resp[‘hits’][0][‘text’] if confidence < self.min_confidence: return None return (confidence, text) class NludbChatbot(NludbChatbotBase): “”” This class is intended to wrap the core functionality with any extra personality, error messages, etc. Also can provide a threshold here below which either no fact is returned or the bot humorously tries to hedge its answer.. “”” def __init__(self, api_key: str, server_name: str): super().__init__(api_key, server_name) def dispatch(self, message: str, username: str = None, roomname: str = None) -> str: message = message.strip() words = re.split(‘\s+’, message) if len(message) == 0 or len(words) == 0: return “””QQ can learn to answer questions. Learn something: `qq fact: Bob’s birthday is in January.` `qq fact: See Jamie about new account setups.` Ask something: `qq when is Bob’s birthday?` `qq who do I see about creating a new account?””” elif bot.arguments == ‘forget everything you know’: return “I don’t yet know how to forget” maybe_fact = self.get_fact(message) if maybe_fact is not None: if len(maybe_fact.strip()) == 0: return “Say `qq fact Some fact: goes here` to learn something” super().dispatch(message, username, roomname) return “{}… got it!”.format(maybe_fact) # It’s a question resp = super().dispatch(message, username, roomname) if resp is None: return “I’m not really sure how to answer that..\n\nAdd new facts with `qq fact: Some fact goes here`.” score, text = resp return text secret = bot.secrets.read(“nludbapikey”) if secret is None: bot.reply(“Please set the `nludbapikey` secret in https://Ab.Bot”) else: # if docs is None: # bot.reply(“I haven’t learned anything yet! Say `qq fact Some fact goes here` to teach me”) user = bot.from_user[‘UserName’] #bot.reply(“{}”.format(bot.conversation_reference)) chatbot = NludbChatbot(secret, “AbbotDemo1”) reply = chatbot.dispatch(bot.arguments.strip()) bot.reply(reply)

Version	Created	Author	Release Notes	Installs
1.0.0	4/7/2021	ted	Initial release. Basic question/answering functionality which should be enough to begin playing.	4

Sign in to add this to Slack

You need an account to install this package with Abbot. Signing in with Slack will create an account if you don’t have one already.

Not using Slack? No worries! Abbot works in Slack, Discord, and Microsoft Teams.

It’s free to get started!

Details

Last updated: Wednesday, 07 April 2021 05:33
Installs: 4 total installs
Language: Python
License: MIT License

Abbot – qq for Slack by Alumstabase

About Abbot

README

Usage

Code

Version History

Details