Admin User Guide · Specialised Modules

Beyond the basics.
Purpose-built
intelligence.

Specialised Modules extends the platform with focused, high-powered capability — a safety timer that watches over lone travellers, facial and vehicle recognition, an AI voice assistant, and a full WhatsApp community intelligence pipeline. Each one activates from preprogrammed buttons on the citizen and responder apps.

🛡️ Virtual Guardian 🔎 FaceMap 🚗 VOI Tracker 🤖 Agent AWE 💬 WhatsApp Intel
5
Active modules
3
Coming soon
8
WhatsApp Intel sub-tabs
3
WhatsApp monitoring methods
Scroll to begin ↓
Overview

Specialised Modules at a glance

Five focused tools, searchable by category, each one a self-contained capability that plugs into the wider platform.

Specialised Modules is the platform's extension layer — capabilities that go beyond core event, responder, and communications management into focused safety, investigation, surveillance, and AI-assisted tooling.

Specialised Modules landing page showing 5 Active badge, category filters for Safety, Investigation, Surveillance, Communication, Tracking, Reporting, Utility, and Active Modules cards for Virtual Guardian, FaceMap, VOI Tracker, Agent AWE, and WhatsApp Intel, plus Coming Soon cards for Neighborhood Watch, Live Tracking Share, and Smart Incident Reporter
Specialised Modules → 5 Active Modules, category filters, and a Coming Soon section previewing what ships next
🛡️
Virtual Guardian
Safety
🔎
FaceMap
Investigation
🚗
VOI Tracker
Surveillance
🤖
Agent AWE
Utility
💬
WhatsApp Intel
Communication

Buttons, not menus. Every specialised module is designed to be triggered from a preprogrammed button on the Citizen App or Responder App home screen — configured in Configuration → Extra Buttons — so field use is a single tap, not a navigation exercise.

Section 02 · Safety

Virtual Guardian

A countdown timer that watches over anyone travelling alone — and alerts the platform and emergency contacts if it isn't cancelled in time.

Virtual Guardian is a manual or route-based safety timer: a citizen starts it before a walk, a lift home, or a meeting with a stranger, and if it expires without being cancelled, the platform and their personal emergency contacts are alerted automatically.

Virtual Guardian screen with a No emergency contacts warning, Trip Details field for where you're going and who you're meeting, Voice Note and Photo buttons, Manual Timer selected, a Timer Duration slider set to 15 minutes with quick options 5m to 2h, and a Start Virtual Guardian button
Manual Timer → trip details, voice note or photo evidence, and a duration slider from 5 minutes to 2 hours
Virtual Guardian Route Timer selected with Driving and Walking mode toggle, an embedded map, a destination address field showing 73 Robin Ave Khayalami Hills Midrand, a Calculate Route Time button, and an estimated travel time countdown of 00:06 with a 20 percent buffer, plus a Start Route Guardian button
Route Timer → pick Driving or Walking, set a destination on the map, and the timer auto-calculates from estimated travel time plus a 20% buffer
  • Manual Timer — set your own countdown from 5 minutes to 2 hours for any situation that needs a safety net.
  • Route Timer — set a destination and travel mode; the app calculates estimated travel time automatically, with a 20% buffer built in.
  • Trip Details — log where you're going, who you're meeting, plus an optional voice note or photo, all attached to the guardian session as evidence.
  • Emergency contacts required — the app warns up front if no emergency contacts are configured, since they're who gets notified if the timer expires.

Silent by design, loud when it matters. Virtual Guardian runs quietly in the background during the trip — but if it isn't cancelled before the countdown hits zero, it escalates immediately to both the platform's control room and the citizen's personal emergency contacts.

Section 03 · Investigation

FaceMap

A person-of-interest facial recognition database — build it once, then scan any face against it in the field.

FaceMap lets responders and admins build a database of persons of interest — missing persons, wanted individuals, persons flagged for questioning — and run instant face searches against it from a phone camera.

FaceMap dashboard showing AI Ready status, 7 Active POIs, 7 Faces Indexed, 1 Sighting, a search bar, Face Search and Add POI buttons, and POI cards for Wayne (Missing Person, Medium priority), POI2 (Wanted for Questioning, Medium), Unknown POI (Wanted for Questioning, High), and Unknown (Other, High)
FaceMap → Active POIs, Faces Indexed, and Sightings counters, with Face Search (scan & match) and Add POI (new person) as the two primary actions
Add Person of Interest form with Take Photo and Upload options, Name/Alias field, Reason dropdown set to Missing Person, Priority dropdown set to Medium, Description, Physical Description, and Case Reference fields, and a Create POI Record button
Add Person of Interest → photo, name/alias, reason (Missing Person, Wanted for Questioning, Other), priority, description, and an optional SAPS case reference
Face Search screen with Take Photo and Upload Photo options to scan a face against the POI database
Face Search → take or upload a photo in the field and instantly check it against every indexed POI

10 POI categories, one searchable database. Every person added to FaceMap is indexed for facial recognition matching immediately — a responder who spots someone matching a description can confirm identity in seconds rather than radioing a description back to base.

Section 04 · Surveillance

VOI Tracker

Vehicle-of-interest identification and tracking — photograph a suspicious vehicle once, and the plate is recognised everywhere after.

VOI Tracker automatically extracts license plates from photos using AI, building a database of suspicious or stolen vehicles that can be searched by plate, make, model, or a scanned photo from the field.

Vehicle of Interest dashboard showing 8 Vehicles, 8 Active, 1 Sighting, 23 Searches, 8 Photos, Report Vehicle, Scan Plate, Search DB, and Refresh buttons, a search bar, and a grid of vehicle photo cards with plates BX54RNGP, JR03TNGP, HF79FRGP, SO81TGP and status badges like Suspicious and Active
VOI Tracker → Vehicles, Active, Sightings, Searches, and Photos counters, with Report Vehicle, Scan Plate, and Search DB as primary actions
Report Vehicle form with Capture Plate and Capture Vehicle photo options, Upload from gallery, and Vehicle Details fields for License Plate, Color, Make, Model, Year, Reason set to Suspicious, Distinctive Markings, Description, and a captured location, with a Report Vehicle of Interest button
Report Vehicle → capture plate and vehicle photos separately, auto-filled make/model/color, reason, and distinctive markings
Search Vehicles screen with Scan Plate, Plate Number, and Description tabs, Camera and Upload options, and a Recent Searches list showing plate searches with match counts and dates
Search Vehicles → scan a plate, type a partial plate number, or search by description, with recent searches and match counts saved

AI does the typing. Capture Plate uses automatic license plate recognition to extract the registration number from a photo — no manual entry required in the field, even from a moving vehicle or a quick glance through a fence.

Section 05 · Utility

Agent AWE

A voice-and-text AI assistant that can action any platform feature hands-free — file a report, check missions, or generate an incident summary, just by asking.

Agent AWE is a conversational AI layer over the entire platform — responders and admins can talk to it naturally, by voice or text, and it files reports, checks assignment status, and generates on-demand reports without touching a single menu.

Agent AWE home screen with a Live Incident banner, a glowing avatar showing Ready status and GPS Active, and quick action buttons File a report, Check my missions, Generate incident report, and Who's on duty
Agent AWE → Ready state with GPS Active, and four quick-start actions: file a report, check missions, generate incident report, who's on duty
Agent AWE avatar glowing green with a Speaking indicator and an audio waveform, showing GPS Active status
Voice mode → hold the mic to speak naturally; the avatar glows and animates while Agent AWE listens and responds
Agent AWE chat showing a response referencing the Alexandra Crime Alerts And Chats group, a View Report link, and an embedded Incident Summary Report card for the last 30 days showing 500 Total Incidents, 419 Resolved, 70 Active, with Email Report and Download buttons
Generate incident report → Agent AWE builds a full Incident Summary Report inline, ready to email or download on the spot
Agent AWE Conversation History panel showing a New Conversation button and a list of past conversations including Generate incident report, Can you pull from our WhatsApp intelligence, Is Alexandra safer after 9pm, and How do I go about reporting a pothole
Conversation History → every past exchange is saved and searchable, from report generation to casual operational questions
Agent AWE fresh conversation screen with a greeting message Howzit Greg! Starting a fresh conversation. What can I help you with?, and the same quick action buttons available
New Conversation → a personalised greeting and the same quick actions, ready for a new request at any time

Natural language, real actions. Questions like "Is Alexandra safer after 9pm?" or "Can you pull from our WhatsApp intelligence?" aren't just answered conversationally — Agent AWE queries the platform's real data (events, reports, WhatsApp Intel) to generate an actual, grounded answer.

Section 06 · Communication

WhatsApp Intel Dashboard

AI-powered monitoring of community WhatsApp groups — classifying, mapping, and auditing every message that comes through.

The WhatsApp Intelligence Dashboard is the day-to-day operating view for community WhatsApp monitoring — six tabs covering Overview, Incidents, Messages, Audit, Map, and Settings, all reading from the same classified message stream.

Overview Incidents Messages Audit Map Settings
WhatsApp Intelligence Dashboard Overview tab showing 2854 Messages, 726 Classified, 105 Incidents, 92 Events Created, 11 Active Groups, 0 Outbound, an Incident Type Breakdown bar chart, and a Severity Distribution chart
Overview → Messages, Classified, Incidents, Events Created, Active Groups, and Outbound counters, plus Incident Type Breakdown and Severity Distribution

Incidents — confirm, dismiss, or escalate

Every AI-classified incident lands here with a confidence score, sentiment tag, and one-click actions — Confirm to validate it, Dismiss if it's a false positive, Escalate for urgent review, or Create Event to push it straight into the main Events system.

Incidents tab showing 2047 incidents, filters for All Types, All Severity, All Status, a search bar, and incident cards including Community Info at 80% confidence about looting events, a Shooting incident at 95% confidence about a death in Sitjwetla, and another Community Info incident, each with Confirm, Dismiss, and Escalate buttons
Incidents → type, severity, and status filters, confidence scores, sentiment tags, and Confirm / Dismiss / Escalate / Create Event actions per incident
Messages tab filtered to Fourways CPF Alert Network showing 288 messages from senders like Sarah Thompson, Nomsa Khumalo, Lerato Dlamini, and Pieter van der Merwe, each classified with a Re-classify button, including a missing child alert and reports of robbery and break-ins
Messages → every raw message per group, classification status, GPS coordinates where available, and a manual Re-classify option
Audit tab showing 4141 total classifications in a Classification Audit Trail, listing entries marked Not an Incident with confidence scores and reasoning, such as a message containing only a TikTok link or a casual greeting
Audit → a full Classification Audit Trail with the AI's stated reasoning for every call, including why a message was ruled Not an Incident
Map tab with a 90 day time range, Heatmap toggle, 291 incidents color-coded by severity, and pinned incidents on a map including a General Alert, a Theft report, and a Missing Person case with full summary details in popups
Map → every geolocated incident plotted and color-coded by severity, with a Heatmap toggle for density view over any time range

Full transparency by design. The Audit tab exists specifically so any classification decision can be reviewed and explained — critical for a system that's making judgment calls about what counts as a safety incident from informal community chat.

Section 07 · Configuration

Intelligence Management Console

The engine room behind WhatsApp Intel — monitoring methods, keyword alerts, live feeds, and the full event-creation pipeline.

Intelligence Management is a deeper, admin-only configuration console reached from the WhatsApp Intelligence Dashboard — 8 tabs controlling exactly how messages are captured, classified, and turned into platform events.

Methods Forwarding Sessions Keyword Radar Feeds Unified Feed Threads Pipeline

Methods — three ways to capture WhatsApp intelligence

The platform supports three independent monitoring methods, each with different risk and setup profiles, and all three can run simultaneously feeding the same unified pipeline.

Intelligence Management Methods tab showing Active Methods 3 of 3, Messages 30 days 2854, Incidents 30 days 105, Pipeline Status Operational, an Admin Forwarding card marked No Risk and Active with features and requirements listed, and an Official Groups API card marked Compliant and Active
Methods → Admin Forwarding (No Risk), Official Groups API (Compliant), and Baileys (High Risk) — 3 of 3 active feeding one pipeline
No Risk
Admin Forwarding
Community admins forward messages from their groups to the AWE business number. Zero risk, works immediately, with self-service REGISTER onboarding and quick report commands.
Compliant
Official Groups API
Meta's official WhatsApp Business API for group messaging — create new monitored groups and invite community members directly, fully compliant with Meta's terms.
High Risk
Baileys (Unofficial)
Connects via the WhatsApp Web protocol to join existing groups for silent monitoring — powerful for joining groups that already exist, but unofficial and carries account ban risk.
Baileys Unofficial method detail showing High Risk and Active badges, 2854 messages, 105 incidents, 3 groups, 9912 total processed, features like silent monitoring and QR code pairing, requirements including a dedicated phone number and accepting risk of account bans, and a Daily Message Volume by Method chart
Baileys detail → silent monitoring, QR pairing, multiple session support — paired with a daily message volume chart showing real throughput

Baileys carries real risk. The unofficial WhatsApp Web protocol is not supported by Meta and can result in account bans — the platform requires a dedicated phone number (never a primary number) and explicit acceptance of that risk before enabling it.

Forwarding & Sessions

Forwarding manages the Admin Forwarding method specifically — registered forwarding admins, their linked communities, and the quick report commands (CRIME, FIRE, MEDICAL, HIJACK, and more) they can text to log an incident instantly. Sessions manages the live connection status for both Baileys and Official Groups.

Forwarding tab showing the Admin Forwarding Method explanation, Registered Forwarding Admins with 2 admins including Test Admin (Verified, active) and Update Test (Verified, Inactive), and a grid of Quick Report Commands like CRIME, ROBBERY, FIRE, MEDICAL, HIJACK, SHOOTING, SUSPICIOUS, MISSING, ASSAULT, GBV, THEFT, and BURGLARY
Forwarding → registered admins per community, verification status, and the full quick report command vocabulary
Sessions tab showing an Unofficial API Warning about account bans, Baileys Sessions with 1 sessions configured, and a Groups session showing connected status with 27625010251, 3 groups, and 2221 messages, with a Disconnect button
Sessions → live connection status per Baileys session, with group count, message throughput, and one-click Disconnect

Keyword Radar — automatic detection across every method

Keyword Radar creates watchlists of specific terms — flagged automatically the instant they appear in any monitored message, across all three monitoring methods simultaneously, with real-time alerts.

Keyword Radar tab showing 2 watchlists active: Immigrant Dilemma (Active, Alerts On) with keywords march on march, genocide, illegal immigrants, and 2 matches; and Weapons Keywords (Active, Alerts On) with keywords gun, firearm, weapon, and 44 matches
Keyword Watchlists → "Weapons Keywords" alone has logged 44 matches; every match is timestamped with a Last Match date

Feeds & Unified Feed

Feeds shows every monitored group side-by-side in a live column view, filterable by source method. Unified Feed flattens all 12,000+ messages across every method into one chronological stream, with the AI's classification reasoning shown inline for messages ruled not to be incidents.

Feeds tab showing multiple group columns side by side: Alexandra Crime Alerts And Chats, Main group Public Safety Community Guardians, Awe Intelligence test group, Durban North Neighbourhood Watch, and Soweto Community Guardians, each showing live messages with severity tags and Push Event indicators
Feeds → every monitored group as a live column, with high-severity messages flagged and pushed to Events automatically
Unified Feed tab showing 12033 messages across all methods, a chronological stream from the Baileys method in Alexandra Crime Alerts And Chats, including messages ruled not incidents with AI reasoning shown inline
Unified Feed → 12,033 messages across every method in one chronological view, with AI reasoning shown for filtered-out messages

Threads & Pipeline

Threads groups related messages into a single incident narrative — a shooting report and the follow-up messages about it become one thread rather than scattered entries — filterable by Active, Monitoring, Resolved, or Stale. Pipeline is where the AI's event-creation behaviour is tuned directly.

Threads tab showing 421 threads, filters for All, Active, Monitoring, Resolved, Stale, and thread cards including Fatal shooting in Sitjwetla Alexandra after fight escalated, Gunshots Heard Unknown Location, Death reported at John Brand Street Alexandra, and Assault Zimbabwean person beaten during march
Threads → 421 tracked incident narratives, each linking related messages to a single Event ID over time
ON
Auto-Events
Medium (3)
Min Severity
418
Active Threads
343
Threads → Events
Pipeline tab showing Event Pipeline Configuration, an Auto-Create Events toggle on, Minimum Severity for Auto-Events set to Medium 3 out of Info Low Medium High Critical, and Incident Types for Auto-Events with many types selected including crime, robbery, burglary, theft, assault, hijacking, shooting, fire, accident, medical, and more
Pipeline → Auto-Create Events toggle, minimum severity threshold (Info through Critical), and which incident types are allowed to auto-create events
Pipeline tab continued showing Notify on Auto-Event toggle, Auto-Register Groups toggle, and Location Intelligence section explaining Shared Locations, AI Text Extraction, and Geocoding via Google Maps for accurate map plotting
Location Intelligence → shared GPS pins, AI text extraction of street names and landmarks, and geocoding via Google Maps for accurate event plotting

Tune the threshold to the community. Setting Minimum Severity to Medium means only genuinely actionable incidents auto-create events — a chattier, lower-risk community can raise this threshold to Critical, while a high-crime area might lower it to Low for maximum coverage.

Section 08

Coming Soon

Three more specialised modules already on the roadmap.

Specialised Modules is an actively growing library. Three additional modules are already visible in the platform, flagged Coming Soon, previewing where the module set is headed next.

👁️
Neighborhood Watch Safety
A community-driven surveillance and coordination tool, extending the safety category alongside Virtual Guardian.
📍
Live Tracking Share Tracking
Share a live location feed with trusted contacts or the control room for an added layer of real-time visibility.
📝
Smart Incident Reporter Reporting
An AI-assisted incident reporting flow, likely extending Agent AWE's natural-language reporting into a dedicated guided tool.

Have a module in mind? If a specialised capability your organisation needs isn't on this list or in the standard module catalogue, chat to our Dev Team — new modules are added to this library as they're built.