Internal DocsModulesAuto Responses

Auto Responses

Auto Responses are a major feature of Architus that lets users pre-configure specific trigger-response pairs for the bot to say whenever it detects the trigger occurring in text chat. This can be useful for server administrators to configure custom commands or otherwise automatic responses.

Triggers

There are three main types of auto response triggers that determine the way text chat messages are scanned and matched to triggers. Note that regardless of the mode, the original triggers will be retained for display on the web dashboard.

  1. No punctuation: Text-based trigger that ignores whitespace and punctuation in trigger
    • Strips all whitespace and punctuation from matching trigger and matching text
  2. Some punctuation: Text-based trigger that ignores all whitespace and punctuation except what appears in the original trigger
    • Strips all whitespace from matching trigger and matching text, but not punctuation that is used in original trigger (excluding mention and emoji fragments)
    • Regex escaping is added for everything
  3. Begins with ^ and ends with $: matches a regex
    • Implement in future release?
    • How to prevent catastrophic backtracking/ReDoS?
    • ^.*aba.*$
      • Makes it opt-in to have “glue” instead of opt-out for better default performance
    • Don’t really need to escape anything

Mode Assignment Heuristic

The above modes will automatically be assigned depending on an educated guess from the backend. To accomplish this, the triggers must be first fragmented to only look at non-emoji/mention components:

Fragments

Original trigger string: <@54435432534524423> hello :) becomes:

[ 
   { 
      "type": "mention",
      "text": "<@54435432534524423>"
   },
   { 
      "type": "fragment",
      "text": " hello :)"
   }
]

Only text fragments will be used for the mode assignment heuristic and the sensitive punctuation pool.

Auto Response Object

Auto Response Structure
FieldTypeDescription
idhoar frostunique hoar frost ID for the auto response
triggerstringoriginal trigger from command/added response
trigger_regexregexderived regular expression from (escaped) trigger text
trigger_punctuationtuplelist of punctuation that a sensitive trigger cares about
responsestringresponse text to use upon invokation
response_astjson stringlexed and parsed response
modestring enumtype of matching mode to use. One of ["naive", "punctuated", "regex"]
author_idsnowflakeuser id of the author
guild_idsnowflakeguild id of the owning guild
countintegernumber of times the auto response has been invoked

Quotas

Quota setting for each non-admin user to have maximum number of auto-response to be able to use.

As a potential reach goal, roles can be used to assign more specific quota amounts within settings. This can be implemented in a future version.

Length Restrictions

Settings for minimum length for trigger and maximum length of response

Consider implementing internal maximum length of trigger restriction to prevent extremely long regular expressions.

Trigger Priority

Have priority be a function of mode and author (admin or not). This should be reflected on the frontend somehow.

  • This is irrelevant if collisions are impossible

UI Context view

Include info about previous conflicts & previous triggers (integration with logs)

Command Conflicting

Check for collisions when trigger regex is generated? potential resource. This option might require a limit on responses per guild depending on how expensive it is to check.


Joseph Azevedo profile
Last modified 4 weeks ago
Edit this page on GitHub
Built with using