Changelog: v3.x

The v3.x line: multi-tenant rewrite, Pulse, Self-trained AutoMod, Bot Health Insurance, Threat Score.

Tap a release to expand it.

Release typeFixFeatureUpdate

v3.8.0: Threat Score (advisory)

2026-05-29

Added

  • Threat Score (advisory). Each user now gets an at-a-glance risk score (Low / Elevated / High) shown in their profile (/info and the right-click View Info menu). It’s transparent and explainable: a time-decayed weighted view of a user’s infraction history (manual warnings count more than AutoMod hits; serious categories like scams/raids/slurs count more; old infractions fade), with a one-line summary of what’s driving it. Advisory only. It never takes any action. It just helps staff see who’s trending risky and decide. Score-driven moderation (with grace periods, soft-touch DM check-ins, server-defined thresholds) is a separate, later phase. Opt-in (slated premium), off by default: enable under /settings → Behavior.
v3.7.0: Bot Health Insurance

2026-05-29

Added

  • Bot Health Insurance. The bot now watches its own footprint so it never quietly becomes the bot a server mutes. If its AutoMod activity looks heavy-handed over a week, it posts a single gentle check-in to staff chat with concrete ways to dial it back (gentler filter pack, softer ladder, trimming custom filters). Read-only: it never changes settings on its own, and it’s self-limiting to at most once every 14 days. On by default; toggle under /settings → Behavior.
v3.6.0: Self-trained AutoMod

2026-05-28

Added

  • Self-trained AutoMod (/automod → Self-training). The bot learns from your moderators, not just preset filters:
    • Watches the content of messages your staff delete or flag via Message Report (opt-in, stored locally for 30 days: message text only, no identities, nothing leaves the host)
    • Learns on demand. Once it has enough samples, the AI clusters the patterns and proposes server-specific AutoMod rules, each with a rationale and the examples it would catch
    • You approve each rule one at a time; accepted rules run through the same backtracking-safety guard as hand-written filters. Nothing is ever auto-enforced: they’re suggestions until you accept them
    • Off by default; toggle on and clear the learning log any time
v3.5.2: /vote now links to the correct directories

2026-05-28

Fixed

  • /vote now links to the correct directories. The “Discord Bot List” link previously pointed at the support server by mistake; it now goes to the actual listing, with dedicated upvote buttons for both Top.gg and Discord Bot List. The newly-approved discord.bots.gg listing is included too.
v3.5.1: Pulse digests carry a gentle vote reminder

2026-05-28

Changed

  • Pulse digests carry an occasional gentle vote reminder: at most once a week, the daily digest adds a single line noting that a quick /vote helps other communities discover Server Assistant. No extra posts, no nagging.
v3.5.0: Pulse: daily server-health digest

2026-05-28

Added

  • Pulse: daily server-health digest (/settings → Pulse). Once a day at a time you choose, Pulse posts a one-embed summary to your staff chat:
    • 24-hour activity (messages, joins/leaves, AutoMod blocks, warnings) with week-over-week deltas vs your 7-day baseline
    • Leading-indicator flags when something’s notable: AutoMod activity is 2× the 7-day average, or join spike above the baseline but below the raid threshold
    • Read-only. Pulse never auto-actions. Just gives staff x-ray vision so they decide what to do.
    • On by default for configured servers: posts to your staff chat at 09:00 server-local time. Toggle off, change the destination channel, or pick a different hour anytime in /settings → Pulse.
v3.4.2: Beta-period perk made explicit

2026-05-21

Changed

  • Beta-period perk made explicit across the site. Servers that add Server Assistant before the premium tier launches receive 12 months of free premium access from launch day. The welcome DM new owners receive now mentions this too: invite the bot now, lock in a year of premium for free.
v3.4: Notification Roles

2026-05-20

Added

  • Notification Roles (/settings → Notifications). Pick which staff roles get @-pinged on three high-signal events:
    • Anti-raid alerts (mass-join detection)
    • Dangerous-action approval requests
    • AutoMod escalations (timeout / kick / ban)

    Up to 10 roles per event type. Leaving a selector empty makes that event silent. Existing behaviour preserved for anyone who doesn’t change anything. AutoMod single-message blocks intentionally don’t ping, to avoid channel noise.

Fixed

  • AutoMod now DMs the user when their message is removed. Previously the message vanished silently with no explanation. The DM names which rule fired, shows the user’s current warning count, and warns about escalation.
  • Duplicate response bug where the bot occasionally replied twice to the same staff message.
v3.3: Weekly setup reminder

2026-05-10

Added

  • Weekly setup reminder: if /setup hasn’t been completed in your server, the bot DMs the server owner once every 7 days with a link to the setup guide. Falls back to the server’s system channel if owner DMs are closed. Stops automatically once setup is complete.
v3.2: Directory stats sync, regex safety net & scheduling fix

2026-05-10

Added

Security

  • AutoMod custom-regex safety net: admin-supplied filter patterns are now rejected at add-time if they contain catastrophic-backtracking constructs (nested quantifiers, alternation-in-repeat, optional-in-repeat). Stops a single bad pattern from impacting the bot’s responsiveness.

Performance

  • Faster user-history search across multi-channel allowlists (channels scanned in parallel).

Fixed

  • /schedule reminders now actually fire. A bug introduced in an earlier release silently prevented scheduled tasks from running. The user-facing commands (schedule in 2 hours: ..., remind at 9am: ...) and slash-command equivalents work as intended.
v3.2 launch: First-steps guide & public documentation site

2026-05-10

Added

  • First-steps guide: after /setup completes, the bot auto-posts a 5-message walkthrough to your staff-chat covering core commands, slash commands, right-click menus, and /settings.
  • Right-click discoverability: /help, the text help command, and the welcome DM now all surface the Apps ▸ Server Assistant context menus (View Info, View Warnings, Message Report).
  • /vote: quick link to vote for the bot on Top.gg.
  • Public documentation site: full Setup, Commands, Features, FAQ, Support, Privacy, Terms, and this Changelog page.

Fixed

  • Duplicate context-menu entries in the Apps submenu.
  • Natural-language command parsing: phrasings like schedule in 3 days: check ... now route correctly.

Changed

  • Slash commands propagate to your server within seconds of an update, not the previous up-to-1-hour delay.
v3.1: Privacy Policy & Terms of Service

2026-05-09

Added

  • Privacy Policy and Terms of Service documents, required for Discord verification submissions.
v3.0: The major rewrite

2026-05-01

The major rewrite. Server Assistant became a Discord bot for any community, not just one specific server.

Added

  • /setup wizard: interactive 3-step config (channels, roles, AI provider).
  • /settings Customisation Hub: 8 sub-wizards for branding, presets, behaviour, new-member features, verification, channel allowlist, role tiers, and notifications.
  • AI image generation (/imagine): supports DALL·E 3, Stable Diffusion, and Pollinations.ai (free fallback).
  • Per-server branding: embed colour, bot nickname, default timezone.
  • Punishment ladder presets: Gentle / Standard / Strict.
  • AutoMod filter packs: Off / Scams only / Scams + slurs / Strict.
  • Anti-raid sensitivity presets: Off / Normal / Strict.
  • Onboarding template gallery: Friendly / Gaming / Professional / Minimal / Custom.
  • DM verification gate: button-based new-member verification.
  • Custom role tiers beyond Owner / Admin / Moderator.
  • Slash commands and right-click context menus.
Browse by version v6.x latest v5.x v4.x v3.x v2.x v1.x

What’s next?