Text anti-spam

Set up in the Mini App#

Open Settings in your private chat with @defendy_bot, select the group under My groups and check its name at the top.

1. Open Anti-spam → Text.
2. Enable the filter and select its strictness: mild, standard or strict.
3. If AI checking is available, enable it and choose the newcomer period. The newcomer-only condition is not editable in this interface.
4. Choose the action. Set its duration for a mute or ban, or the warning weight for a warning.

Switches and selectors save when changed. In a separate form with a Save button, save before leaving. If a request fails, reopen the section to check its actual values.

Tip: change one option at a time and verify the result in the group so false positives are easy to spot.

In short#

The current chain checks links before text anti-spam, followed by stop words and profanity. Forwarded messages have a forward check first. The chain stops at the first matching filter. Try the settings in a separate test group first: a warning can accompany message deletion and add points toward another consequence.

How it works#

1. The local detector checks message patterns with the selected strictness.
2. If AI checking is enabled, eligible new members receive an additional check.
3. On a match, Defendy deletes the message if configured and applies the selected action or warning weight.

Every setting explained#

Enable filter#

Default: On.

Turns text spam detection on for the selected group.

Strictness#

Default: Standard.

Strictness changes the fallback threshold for uncertain local-model scores: 0.65 for Lenient, 0.55 for Standard and 0.45 for Strict. A lower threshold can trigger more often, including on legitimate messages; it is not an accuracy guarantee. Built-in pattern matches and confident local-model decisions bypass these thresholds, while eligible messages may be decided by AI. See the anti-spam overview for eligibility, provider requirements and fallback behavior.

AI verification#

Default: Off.

Adds AI verification for suspicious messages from new members.

New members only#

Default: On.

Locked in the UI: trusted members stay on local checks only.

New member window#

Default: 7 days.

The window is counted from Defendy’s first stored observation used by this check, not a verified Telegram join time. When that record is absent, even an existing group member can be treated as new.

Message deletion is a separate consequence#

Schema default: On. A particular group’s saved value may differ.

Deletion is separate from the action applied to the member. The current Anti-spam → Text screen has no dedicated deletion switch, so choosing “nothing” or “warn” does not mean that the message stays. Verify the actual outcome in a test group; do not treat the screen as a consequence-free preview mode.

Action#

Default: None.

What Defendy does after a violation: nothing, warn, notify, mute, kick or ban depending on the feature.

Action duration#

Initial value: 1 hour (3,600 seconds).

The mute or ban duration, when that action supports a time limit. Where supported, zero means no expiry. Check the duration again after changing the action.

Warning weight#

Default: 1.

How many warning points this violation adds to the warning system.

Practical example#

In a separate test group, select standard strictness, action “warn” and weight 1. If AI verification is available, select a seven-day newcomer window. Test both allowed and unwanted messages: warnings do not disable deletion, and accumulated points may trigger another punishment. Check warning thresholds and actual outcomes before applying the settings to a working community.

Recommendations#

• Use standard strictness first, then adjust after reviewing real messages.
• AI verification adds latency and should not replace deterministic filters.

Troubleshooting#

• Check that Defendy is still an administrator in the group.
• Refresh the settings panel and make sure you are editing the correct group.
• If a setting is saved but does not work, check the bot permissions in Telegram.
• If logging is connected and the relevant category is enabled, check the event record. The log cannot show events that were not recorded.

Check the duration after changing an action#

Changing the punishment type can reset its duration to 0. Where supported, zero means an indefinite action, not a disabled punishment. A temporary mute lasts at least 60 seconds. Recheck the duration before leaving the form.

When the filter is actually active#

The schema’s initial switch value is on, but the filter does not run if the group has no module settings. Check the actual state in the Mini App and test it; adding the bot alone is not sufficient proof that protection is configured.