worldbuilding / kaztron Goto Github PK

• Current behaviour by design: A field value in excess of 1024 characters, from a spotlight application, is truncated at 1024 characters (incl. an added ellipsis).
• Maybe possible to just add more embed fields with the overflow text, instead of truncating?
• Would need to limit this: max 25 fields, 6000 character limit (#4), may require multi-message overflow too
• Limit also in case input data is unreasonable

Feasibility? Possible to automatically convert the msg attachment to a URL embedded in the modnote?

rotation of kaztron statuses, add "playing with your hats" (requested by Derpquake on worldbuilding)

User input is very iffy for Discord ID. Improve robustness:

• Allow <@123456789012345678>, @123456789012345678, etc. - maybe look for a digits-only substring for maximum flexibility
• Detect completely wrong data - is this handled adequately currently? Verify that nicknames, discriminators, etc. are rejected

e.g.

in showcase, title is

Spotlight Application #6
username#3272

internal/mgmt commands, title is

Spotlight Application #6
username#3272

2018-02-23

[21:03] BOTBotTest9000: [ERROR] While executing [2018-02-24 02:03:58] <#bot_test:AntimatterNuke#7807> '.spotlight join'

Discord API error FORBIDDEN (status code: 403): Cannot send messages to this user

See log for details

• Automatic tempban (mute) timing
• Automatic detection of tempban (mute) evasion

• .notes group rem - if user is not in a group, should give an error message, not 'ungrouped'
• .notes group add - if two users are already grouped, should give an error message, not 'grouped'

Both operations are idempotent - functionally no change, purely a usability issue

• sprint (timedeltas)
• wordfilter (timestamp)
• modnotes
• spotlight (dates in queues)

Our next host is User#1234, presenting their project, Project Title!

Welcome, User#1234!

Username is not a mention in these messages

Priority TBD

• Muting, banning, channel lockdown, etc. - convenience tools/commands for moderators

[2018-02-11 16:11:18,241] (DEBUG) websockets.protocol: client << Frame(fin=True, opcode=8, data=b'\x03\xe8') [in /home/marcalexc/.virtualenvs/kaztron/lib/python3.6/site-packages/websockets/protocol.py:548]
[2018-02-11 16:11:18,241] (DEBUG) websockets.protocol: client >> Frame(fin=True, opcode=8, data=b'\x03\xe8') [in /home/marcalexc/.virtualenvs/kaztron/lib/python3.6/site-packages/websockets/protocol.py:564]
[2018-02-11 16:11:18,244] (INFO) kaztron.bootstrap: Closing event loop... [in /home/marcalexc/data/sw/KazTron/KazTron.py:64]
[2018-02-11 16:11:18,245] (INFO) kaztron.bootstrap: Exiting. [in /home/marcalexc/data/sw/KazTron/KazTron.py:66]

No notable traffic prior to this - normal message send/receives, user status changes, at least 60s prior to this.

Appears to be the remote server closing the websocket?

Per discussion with Yazuki:

• whenever changing or displaying the current spotlight app
• Warn (as a separate message, to ctx.message.channel), if the selected user is not on the server

• Challenges: intuitive command syntax?
• On one hand, .filter rem <index or page> excludes numerical match (possible scenario: bot spam consistently uses a very identifiable string; could be dangerous because if allowed on add, a rem could accidentally hit another rule instead of the numerical match)
• Other syntaxes like .filter rem index=x would disallow an index=x string, which is unlikely enough that this might be OK. However, it's unintuitive
• Separate syntax like .filter rem vs. .filter rnum - unambiguous, but pollutes namespace, more commands to remember - slightly less user friendly?

Affected: v1.3b2

From Grine_

2018-02-28 00:02:47

A bot-managed “On Duty” role, that can be used to notify active moderators without pinging atHere or atDiscordTeam. Mods could set themselves as on or off duty at will.

• only role mgmt command, manually managed by mods

@Yazuki (屮ﾟДﾟ)屮

2018-03-08 00:18:47

fix kaztron so it doesn't think .___. is a command and not a smiley

Possible solution: Assume anything that matches ^\.\W is not a command? Or anything not containing word characters at all (any emoticons that contain word chars but start with a period)?

• Allow specifying and storing start time in spotlight queue - recordkeeping only, no automatic bot action based on date
• queue ordering should become chronological by start time; breaking UX changes are OK

TBD which logger names to limit... mostly library dependencies e.g. websockets.protocol is horrendously verbose. We want to reduce this verbosity without losing out on debug logs from KazTron itself or other useful logs.

Rationale: 600MB log in 1 week

Currently only implemented in logs and terminal:

Logged in as KazTron Test - Lao (id:412447001142231041)
KazTron version 1.3b2
discord.py version 0.16.12

Affected version: 1.3 beta-2

e.g.

[15:01] Laogeobunny: .notes restore 0001
[15:01] BOTKazTron Lao: Only mods can use that command.

Expected: "Only admins can use that command."

• Un-hardcode the spotlight URL

• Current situation: commands are logged by each command, manually - a lot of copy-pasted logging code
• Enhancement: Create a global pre-command hook
• May depend on Discord.py 1.0.0 (unreleased)? Mentioned as a new feature in the transition document

Several cogs are dependent on on_ready running prior to commands being available (loading data from servers, etc.)

A global or optional cog-level check to prevent commands being run before on_ready would avoid bugs due to this (including some destructive ones - e.g. if the bot hasn't loaded state.json configs and a command attempts to write to these structures, it can delete data)

May be related to the proposed refactors of #48

[2018-03-02 06:38:27,215] (DEBUG) kaztron.cog.modtools: wb: random image = 7 [in /kaztron-prod/KazTron/kaztron/cog/modtools.py:109]
[2018-03-02 06:38:27,215] (WARNING) kaztron.cog.modtools: wb: Invalid index: [2018-03-02 06:38:27] <#worldbuilding:Laogeobunny#4569> '.wb' [in /kaztron-prod/KazTron/kaztron/cog/modtools.py:114]

len(images) == 7, range should be 0-6 but here rolled 7

Various issues with voice features:

• Fails to retrieve role (get_named_role call does not pass server)
• Fails to identify transition from a listed to non-listed voice channel, and remove voice role

From: Grine_

2018-03-01 02:32:53

Detect users leaving and report to #bot-output

Currently, most cogs manually call get_kaztron_config() and store a reference in an attribute, as well as similar calls for the state.json config and any cog-specific config files. For the latter two, many of them also tend to have

Can this be generalised to reduce repetition?

• Base class that loads and provides the config object attribute, as well as direct accessors of some sort for the cog-specific and general (core/discord) sections of the config files?
• on_ready status/precondition on all commands? (pre-command hook - can it cancel command?)
• Abstract base class with deserialise/serialise hooks to load/store data in config?

• e.g. .notes @user alias add <alias> instead of .notes alias add @user <alias>
• 1.4 major usability rework?
• Determine scope: full changeover (training problems since 1.3 release is public)? Or add as an alternative (some possible disambiguation problem with 1.3 syntax)?
• Custom parser required: effort vs benefit?

Cache issue?

Need logs.

• Allow users to add/remove themselves from specified roles
• Specify command/role pairs by config
• Specify command/role pairs programmatically
• Convert spotlight and sprint role management