Discord AutoFAQ

This bot written in python makes use of machine learning to detect messages in a Discord channel which can be answered with a predefined FAQ. In order to achieve this, the bot uses BERT for natural language processing (NLP) to calculate sentence similarities between FAQ triggers and new posted messages. Now, if the highest calculated metric for a message exceeds a threshold, the bot automatically reacts with a predefined answer corresponding to the trigger.

If you are interested in BERT, here are 2 links for you:

• Blog post: BERT For Measuring Text Similarity
• Releasing paper: Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

Setup

As the administrator, you first have to create an FAQ topic by using /faq_enable. Then, you can create FAQ entries with /faq_add and fill them with messages which should be automatically answered with your FAQ.

Commands

Command	Description	Permission
/faq_help	Displays the most relevant things to know for staff members.	moderate_members
/faq_enable	Enables AutoFAQ for a Discord channel (and creates an FAQ topic).	moderate_members
/faq_disable	Disables AutoFAQ for a Discord channel.	moderate_members
/faq	Shows all FAQ entries for a specific topic.	moderate_members
/faq_add	Creates an FAQ entry for a specific topic.	moderate_members
/faq_edit	Edits an FAQ entry for a specific topic.	moderate_members
/faq_delete	Deletes an FAQ entry for a specific topic.	administrator
/faq_reload	Reloads every classifier for each topic.	moderate_members

In-Chat Commands

Command	With reference to another message	Description	Permission
@AutoFAQ <faq-abbreviation>	Yes	Posts the answer of your FAQ entry to the referenced message and adds the referenced message to your dataset.	moderate_members
@AutoFAQ <faq-abbreviation>	No	Posts the answer of your FAQ entry.	moderate_members
@AutoFAQ ignore	Yes	Adds the referenced message to the ignore-dataset and deletes any old FAQ response referenced to that message.	moderate_members

Requirements

• python installed
• pip
• a server to run this bot
  • The bot requires at most times ~350MB RAM to load the classifier and
  • ~3GB disk space to be able to install required libraries.

Bot Installation

1. Open the Discord Developer Portal
2. Create a new application
3. Add a bot to your application
4. Enable the message content intent (used for automatically checking messages)
5. Create an OAuth2 link with following permissions and the scope bot
   • Read Messages/View Channels
   • Send Messages
   • Read Message History
   • Add Reactions
6. Connect your bot to your server

Run your Bot

1. Download the source code of this bot
2. Install all requirements of the bot with pip install -r requirements.txt
3. Run the bot with python bot.py