Features
Everything a moderation tool needs, and nothing that pretends to be more than it is.
Sedge is a rule engine, two review queues, an activity model and an audit trail. That is the whole product, and each part is built to be checked by a person.
The signature screen
Every automated action carries its reasoning.
Not “removed for spam”. The score, each rule that contributed, the evidence numbers behind it, the threshold in force, the action taken, and whether it was carried out or only simulated.
The same breakdown appears on the event, on the review case, and in the audit trail — so somebody who was not there can read it months later and understand what happened.
| Rule | Running total | Contribution |
|---|---|---|
Repeated content duplicate.repeated The same message had already been posted twice in the last hour.
| 30 | +30 |
Repeated links link.repeated The same link appeared in each copy.
| 60 | +30 |
Group invite spam link.invite_spam An invite link to another WhatsApp group, posted by a member who is not an admin.
| 95 | +35 |
Message flooding flood.burst Six messages inside the five-second window.
| 115 | +20 |
Previous warning history.prior_warning Warned nine days ago; the weight has decayed from 30.
| 135 | +20 |
| Total | threshold 100 | 135 |
The score reached 135 against a threshold of 100. The group is in monitor mode, so the member was removed only on paper — nothing changed in WhatsApp.
Detection
What the rules look for.
Weighted spam scoring
Signals add up rather than each acting alone. Bands you set decide what a total does.
Flood and media flooding
Rolling windows on messages, images, stickers, voice notes and emoji-only runs.
Repeated content
Near-duplicate matching that survives reordering, padding, accents and digit-for-letter swaps.
Link protection
Allow and block lists, unknown-domain policy, shortener expansion, invite-link detection and repeat counting.
Mass mentions and new members
Mention limits per message and per window, plus stricter handling during a member’s first hours.
Posting patterns
Machine-like regularity and the same content across many groups. Posting often is never enough on its own.
Control
What a person stays in charge of.
Two review queues
Spam and inactivity are separate workflows with different evidence and different outcomes.
Inactive member management
Warnings on a schedule, then removal into a queue. Measured only from observed activity.
Monitor, enforce, paused
Per group. Monitor is the default and switching to enforce takes an explicit confirmation.
Policy inheritance
Platform, organisation, WhatsApp account, community, group. Each value shows which layer set it.
Audit trail
Actor, action, object, old value, new value, time and correlation id — for every change, exportable as CSV.
Reports
Daily and weekly moderation, activity, inactive members, top spam sources, review outcomes, admin actions.
Running it
The parts that keep it running without you.
Connection health
Connected, connecting, scan needed, degraded, reconnecting, session expired, disconnected. Reconnection is automatic where it can be.
Roles that mean something
Owner, admin, moderator, reviewer, billing and read-only, gated on permissions rather than role names.
API and webhooks
Scoped API keys and signed webhook deliveries with retries, a delivery log and a dead-letter state.
See it on your own groups first.
Monitor mode runs the identical code path as enforcement — the only difference is whether the action is carried out. A week of it tells you more than any demo.