| Commit message (Collapse) | Author | Age | Lines |
|---|
| ... | |
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The filterlist describe command is changed to include the default settings it contains.
The filters group can now take a filter ID, and it will display embed detailing the filter data and settings. The mechanic is similar to how individual infractions can be displayed with `!infraction <id>`.
- To be able to quickly find whether the filter with the provided ID is in a specific list, the data structure was changed to a dictionary.
- To be able to mark which settings have the default values and which are overrides, resolving the full actions of a filter is deferred to when the filter is actually triggered. This wasn't possible in the beginning of development, but now that each filterlist can resolve the action to be taken with its own internal logic, it is.
- Some attribute names of SettingsEntry subclasses were changed to match the name under which they're stored in the DB. This allows displaying the settings with the names that need to be used to edit them.
- Each filterlist now contains all settings, even if they're empty, so that they can be displayed. While this is slightly less efficient, I considered it too negligible to make displaying the settings messier than it already is.
- Some additional refactoring in the cog was done to avoid code repetition.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The system is fairly complex, and to avoid having to delve into the source code, this commit adds several commands to show descriptions for various components of the system:
- Filters
- Filter lists
- Settings (both from the DB schema and filter-type-specific)
The names that can be described are the ones which are actually being used in the system. So if no tokens list was loaded, it can't be described even if there is an implementation for it.
If no name is relayed to the commands, the list of options will be shown instead.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Pydantic is added to the bot's dependencies, allowing easily setting up settings that are specific to a filter type (e.g for domains).
The data for those settings is loaded from the "additional_field" JSON field in the Filter table.
This commit specifically adds an option to match a domain exactly, without matching subdomains.
There are several benefits to using Pydantic:
- Built-in serialization to, and de-serialization from JSON, which is useful for reading and writing to the DB.
- Type validation on initialization.
- By default ignores extra fields when de-serializing, which can prevent things breaking if the extra fields change.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The domain filtering works very similarly to the token filtering, and the domain matching itself is based on the implementation in the old system.
The deletion setting is accessed explicitly in the domain filter in order to allow DMing the user the domain the message got deleted for.
This is fine, since practical uses are more important than the theory this system was designed with. Of course, breaking the design should still be avoided whenever possible.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
Previously the completed arg would be appended to the end of the message, which didn't work if the missing argument wasn't the last one (for example, a value was given to a following argument because of a failed conversion for previous ones).
The select now employs a different strategy of storing the provided args, and the position where the new are should be inserted.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This commit adds the invite filtering implementation to the new system.
This also fixes an issue with the implementation of the extension filtering, where there was no way to tell the bot to ignore a user when they posted a non-whitelisted file extension, since there's no relevant filter in this scenario.
Instead the extensions and invites filters now use the whitelist validation defaults to dictate when filtering should be done at all. For example, if the list validations are to ignore Helpers, then no invite filtering will occur no matter what.
The meaning of this is that the system is somewhat less configurable, because settings overrides in filters belonging to a whitelist are meaningless.
Additionally this commit contains the following fixes:
- If the user tries to show the filters in a list which doesn't exist, it says it doesn't exist instead of saying there are 0 filters.
- The filter context content is `Union[str, set]` instead of `Union[str, set[str]]`.
- An empty embed will no longer be created when `dm_embed` is empty.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This commmit migrates the AntiMalware cog to a new filter list which goes over a message's attachments.
Some changes were needed to accomodate the new list, primarily what a filter list returns for a given context:
Instead of returning a list of filters, it will return the action itself that should be taken. This adds the flexibility of not needing existing filters to dictate the action. For example, in the case of the extensions list, an action should be taken when filters were *not* triggered. Or more precisely, when not all attachment extensions are whitelisted. Therefore, the action in that case is dictated by the filter list (stored as the list's default actions).
Additionally each filter list can now return its own message for the alert embed, instead of the cog formatting it according to the filters raised. Because again, an action might be taken without any deny filters being triggered. This is going to be especially relevant for the invites list.
Additionally, the infraction_and_notification action now doesn't redirect the notification to the context channel when the DM fails, since this can be incredibly noisy in cases of spam. If we want this functionality, a more suitable solution should be found.
|
| | | | |
| | | |
| | | |
| | | |
| | | | |
The channel scope settings were changed to accomodate strings. That means that if a string is specified, the bot will look whether the context channel's name matches. If it's a number, it will match the ID.
Accordingly the same changed was applied to the bypass roles and pings settings: if it's a non-numeric string, it will look for a role with that name.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
For legacy purposes, separate command groups were re-added for blacklists and whitelists.
There's a new command group for filters.
Not specifying a list type for the `filter list` command will cause the bot to try to infer whether there's only one kind of list,
for example `!filter list tokens` will pull up the blacklist since there's no whitelist.
If a required field is missing, the user will be prompted to complete it from a selection.
Some level of flexibility was added for the list type in the `filter list` command. For example, a list type can be "DENY", but it will also match "blacklist", "denylist", "deny", "denied", "blacklisted" etc.
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This commit provides the basis of the new filtering system:
- The filtering cog consists of several filter lists loaded from the database (filtering.py).
- Each filter list contains a list of filters, which are run in response to events (message posting, reaction, thread creation). Each filter list may choose to respond to different events (the subscribe method in filtering.py).
- Each filter has settings (settings.py) which decide when it is going to be run (e.g it might be disabled in a specific channel), and what will happen if it triggers (e.g delete the offending message).
- Not every filter has a value for every setting (the _settings_types package) . It will use the default settings specified by its filter list as a fallback.
- Since each filter might have a different effect when triggered, we must check all relevant filters even if we found a triggered filter already, unlike in the old system.
- Two triggered filters may specify different values for the same setting, therefore each entry has a rule for combining two different values (the __or__ method in each file in _settings_types).
To avoid having to prefix each file with an underscore (or the bot will try to load it as a cog), the loading script was changed to ignore packages with names starting with an underscore.
Alert sending is done via a webhook so that several embeds can be sent in the same message (will be useful for example for guild invite alerts).
Filter lists and setting entries classes are loaded dynamically from their respective packages.
In order to be able to test the new features, this commit also includes a migration of the regex-based filtering.
|
| |/ / /
| | |
| | |
| | | |
Tests and dependent functionality in other extensions will be re-added later on.
|
| |\ \ \
| | | |
| | | | |
Add revival of code role to self-assignable
|
| |/ / / |
|
| |\ \ \
| | | |
| | | | |
Fix wrong var name in regex tag
|
| |/ / / |
|
| |/ /
| |
| | |
Also added aliases for the f-string tag and the "minusmpip"
|
| | | |
|
| | |
| |
| | |
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
|
| | |
| |
| | |
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
|
| | |
| |
| |
| | |
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
Co-authored-by: Xithrius <[email protected]>
|
| |\ \
| | |
| | | |
Add tag to use `python -m pip` to install packages
|
| |/ / |
|
| |\ \ |
|
| |/ /
| |
| |
| |
| | |
"Discord" is a proper noun, and should be capitalized.
Signed-off-by: Bradley Reynolds <[email protected]>
|
| |\ \
| | |
| | | |
Improve help of eval and timeit
|
| | |\ \
| |/ /
|/| | |
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
python-discord/always-prepend-guild-name-to-guild-invite-filters
Enhance guild_invite based autobans
|
| | |\ \ \
| |/ / /
|/| | | |
|
| |\ \ \ \ |
|
| |/ / / / |
|
| |\ \ \ \
| | | | |
| | | | | |
Ignore auto mod notificaitons
|
| | | | | |
| | | | |
| | | | |
| | | | | |
This can be unpinned once https://github.com/SebastiaanZ/async-rediscache/pull/18 is merged
|
| | | | | | |
|
| | | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
When a message triggers an auto_moderation_action notification, Discord re-writes the author field for the system message to look as if it's from the original author.
This means those messages counted towards our anti-spam filter. This both triggered the filter too early, and also resulted in those auto mod messages being deleted when the filter was hit.
|
| |/ / / / |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Some guild invites are autoban filters, which require the mod to set a comment which includes [autoban].
Having the guild name in the comment is still useful when reviewing filter list, so prepend it to the set comment in case some mod forgets.
Co-authored-by: Alex Noel <[email protected]>
|
| | | | | |
|
| | |/ /
|/| | |
|
| |\ \ \ |
|
| | |\ \ \
| |/ / /
|/| | | |
|
| |\ \ \ \
| |_|/ /
|/| | | |
|
| | | | | |
|
| |/ / / |
|
| |/ /
| |
| |
| |
| |
| |
| | |
Due to various breaking changes in the discord.py library the views of the help embed stopped working properly.
- Use `add_item` in `CommandView`. This was done in GroupView but was apparently missed in `CommandView`, and the former (hacky) method no longer works.
- Instead of editing the help message normally (which required to then defer the response), the message edit is made to be the response itself. It seems like the callback would automatically defer the response if none was made, but that no longer happens.
|
| |\ \
| | |
| | | |
Only check for autoban filters if reason key is present and not None
|
| | |\ \
| |/ /
|/| | |
|
mbaruh
ChrisJL
Cam Caswell
Hassan Abouelela
Boris Muratov
Janine vN
dependabot[bot]
D0rs4n
Bradley Reynolds
Chris Lovering
Numerlor