MirAI User Manual

1. Overview

MirAI is an intelligent video surveillance and AI analytics platform. It consists of a high-performance Rust-based server (AI inference, stream management, data storage) and a Qt C++ desktop client (live monitoring, incident management, configuration).

Key Features

• Real-Time Video Streaming — Low-latency RTSP via GStreamer with hardware-accelerated decoding. Supports up to 100 cameras per server.
• AI Event Detection — Pipeline-based detection for Flame (fire), Smoke, Falldown, and Intrusion with configurable ROI/RONI zones and confidence thresholds.
• Multi-Camera Grid — Flexible grid (1×1 through 10×10, up to 100 feeds) that automatically expands or shrinks based on the number of cameras placed. Supports drag-and-drop reordering and camera groups.
• Camera Groups & AI Pipelines — Organize cameras into groups. Assign AI pipelines per group.
• Incident Recording & Audit — Automatic recording with a pre-event buffer. Audit workflow (True/False), multi-format download (MP4, AVI, PNG, JPG).
• Notification System — IF-THEN rule engine with Email, Webhook, HTTP Endpoint, or SNMP delivery. Global rate limiting and quiet hours.
• Multi-Server Support — Connect to multiple servers simultaneously, each in its own tab.
• Role-Based Access — Super_Admin, Admin, and User roles with granular permissions.
• System Management — Storage policies, scheduled backup/restore, audit trail, log viewer, automatic updates.
• Localization — English, Korean, and Japanese. Configurable timezone and date format.

Main Screen Layout

System Requirements

User Roles

2. Getting Started

Login

1. Launch MirAI — Double-click the desktop icon.
2. Enter Server IP — Type the server's IP address (e.g., 192.168.1.100). Default port: 7878 (HTTPS).
3. Enter ID and Password — Use credentials from your system administrator.
4. Check "Remember me" (optional) — Enables automatic login next time.
5. Click "Log In" — If the server is starting up, a "Waiting for server..." dialog appears.

First-Time Setup

After your first login, complete these steps in order:

1. Change the default password → Settings → Users → Edit admin account
2. Activate license → Settings → License → enter key or upload file
3. Add cameras → Settings → Server and Camera → + Add Camera
4. Create camera groups → Camera Groups tab → + Create Group
5. Assign AI pipeline to group → Group Details → Configure AI Pipeline
6. Enable detection per camera → Live View → Select camera → AI Settings → enable types

Adding Cameras

Automatic Discovery (ONVIF)

1. Go to Settings → Server and Camera → + Add Camera
2. Select Automatic mode
3. Click Search — cameras discovered on the network appear in a list
4. Check the cameras to add
5. Enter camera credentials (ID/password) if required
6. Click Add

Manual (RTSP URL)

1. Select Manual mode
2. Enter the camera name and full RTSP URL: rtsp://user:pass@192.168.1.50:554/stream1
3. Enter camera credentials (ID/password) if required
4. Click Add

Camera Groups & AI Pipelines

Groups organize cameras and assign shared AI processing. All cameras in a group share the same pipeline configuration.

1. Create group: Server and Camera → Camera Groups → + Create Group → enter name → Create
2. Assign cameras: Group detail → + Assign Camera → check cameras → Assign
3. Assign pipeline: Group Details → Configure AI Pipeline → select pipeline → Add
4. Enable per-camera detection: Live View → Select camera → AI Settings → enable Flame/Smoke/Falldown/Intrusion → Save

3. Live View

Grid Controls

Adding cameras to the grid

• Drag a camera from Server List → drop onto empty grid cell
• Drag a server → fills grid with all cameras from that server
• Drag a camera group → fills grid with all cameras in the group
• Rearrange cameras by dragging between cells

If you re-add a camera that is already on the grid, MirAI highlights the existing tile instead of creating a duplicate.

Grid navigation

The grid size adjusts automatically to the number of cameras placed (for example, adding a camera expands a 1×2 layout to 1×3). There is no separate zoom control. While you drag, a live preview highlights the target cell and shows the resulting grid size (for example, "3 × 2").

Camera Hover Menu

Hover over any camera tile to reveal the control bar:

Notifications Panel (Right Sidebar)

When AI detects an event:

1. A notification card appears in the right panel showing: event type icon, camera name, server, timestamp, and thumbnail
2. Click a notification to jump directly to the incident detail view
3. Click "Clear All" to dismiss all notifications
4. Use the mute button (top-right, speaker icon) to silence alert sounds

Panel behavior: Hovering over the panel pins your current scroll position while new cards continue to arrive, so the list does not jump under your cursor. The panel keeps up to 50 cards (the oldest are pruned automatically). When the panel is collapsed, the reopen icon pulses red whenever a new alert arrives.

Alert sounds: Each event type plays a distinct sound — Fire and Smoke → Alarm, Intrusion → Siren, Falldown → Alert, others → Beep. Sounds are debounced (about 2 seconds) and can be turned off per notification rule using the rule's "play sound" toggle. Overall volume and muting are controlled in General Settings (Alert Volume / Mute All Alerts).

Server Management

Server List Tree Structure

The tree uses presence colors and per-camera status icons:

• Presence color: a camera's name is shown in bright white when it is currently placed on the grid, and grey when it is not.
• AI chip — purple when AI detection is active for that camera.
• Notification bell — yellow when at least one notification rule targets that camera.
• Recording dot — red while recording, yellow when pending, orange on error.

Right-click a camera group to Expand All or Collapse All. You can also drag a camera onto another group to reassign it (a confirmation prompt appears first).

Server Context Menu (right-click a server)

• Server Info — View CPU and GPU product names
• Add Camera — Open the Add Camera dialog
• Test Camera — Open a test video window to verify the stream works
• View Logs — Open Incident Log for this server

Camera Context Menu (right-click a camera)

• Camera Settings — Edit camera name, resolution, profile
• View Logs — View incidents for this camera only
• AI Settings — Configure AI detection for this camera
• View Flow — See the AI pipeline flow diagram (read-only)

Every context menu includes a Show Camera IDs toggle to turn camera ID badges on or off in the grid.

Adding a Server

1. Right-click in the Server List area → Add Server
2. Enter the server's IP address, User ID, and Password
3. Click Test to verify the connection
4. Click Add
5. The new server appears as a new tab at the top

Server Info Dialog

Right-click server → Server Info shows:

• Server Name — editable only on the server you are logged in to and only with Admin/Super_Admin privilege; otherwise it is read-only and the Save button is hidden.
• CPU and GPU product names
• Remove Server button — disconnect from this server

Working with Multiple Servers

MirAI can monitor several servers from one client. The server you sign in to is your login server; any server you add (above) is an added server and appears as its own tab at the top — switch servers by clicking their tabs.

On an added server you can monitor everything: place its cameras on the grid, watch live video, browse its Server List tree, open its Incident Log, and see its AI events — each camera shows the AI event types defined on its own server.

Added servers are view-only. Camera grouping, AI configuration, notification rules, user management, and most settings can be changed only on the server you are logged in to; to configure an added server, log into it directly. (On an added server the Server Name and similar fields are read-only and group drag-and-drop is disabled.)

To stop monitoring an added server, right-click it → Server Info → Remove Server. This only disconnects your client; it does not change the server itself.

Camera Status Indicators

The following states are shown on each camera tile in the Live View grid:

When an AI event fires on a camera, its grid tile briefly flashes a colored border in that event's color — the same color as the event's detection box (red as a fallback). This Screen Flash on Alerts behavior is on by default and can be toggled in General Settings.

Event icons & colors: Each AI event type has an icon and a color set on its event output in the pipeline, and the client uses this as the single source of truth. The event chips in AI Settings and the event labels in the Incident Log show this icon and color, and the alert flash above uses the same color.

4. Settings & Configuration

Access via the Settings button in the top navigation bar. The left sidebar shows setting categories.

General Settings

Available to all user roles. Configure appearance and preferences:

Server and Camera

All Cameras Tab

• Table view: camera name, IP, status (online/offline), group, AI detection status, Alarm (AI Detection Status = whether AI detection is enabled for that camera)
• Search: Filter by camera name or IP
• Filters: Status dropdown (Online/Offline/All), Group dropdown
• Batch operations: Check multiple cameras → Assign to Group or Delete
• Edit/Delete: Per-camera action buttons

Camera Groups Tab

• Group cards showing: name, assigned pipeline, camera count, AI count, alarm count
• Create Group: + Create Group → name → Create
• Assign cameras: Group detail → + Assign Camera → check → Assign
• Configure AI Pipeline: Select pipeline from dropdown → Add

AI Configuration

Access: Settings → AI Configuration. Both Admin and Super_Admin can manage AI models. Only Super_Admin can create/modify pipelines (Admin can view pipelines). For an Admin, the pipeline create, edit, and delete controls appear disabled (greyed out, with tooltips explaining that only Super_Admin can modify pipelines).

AI Model Hub

• View model cards: type badge (Detector/Classifier/Embedder/Segmenter), version, name, description
• Search & filter: Search models and filter by type (All/Detector/Classifier/Embedder/Segmenter)
• Upload model: + Add New AI Model → upload weights (.onnx/.pt/.enc/.tflite) and manifest (.toml) → Save
• Configure: The model card's [Configure] button opens the version dialog. It provides version history, in-place replacement of the model engine/.toml, uploading a new version, and editing the description (the model name is immutable). Changes apply automatically; a server restart is recommended for structural changes. If you change a model that is used by an active pipeline, a restart-recommended prompt is shown.
• Delete: Only models that are not in use can be deleted. Models referenced by an active pipeline cannot be deleted.

AI Pipeline Builder

Per-Camera AI Settings

In Live View, right-click a camera → AI Settings:

• Turn the AI function ON → select the event types to detect: Flame, Smoke, Falldown, Intrusion
• Defined Rules: Expand each event output to view and edit its detection rules in a modal (the pipeline's event-output options, including the confidence threshold). A lower threshold is more sensitive; a higher threshold reduces false alarms.
• ROI: Draw polygon — only detect INSIDE this area
• RONI: Draw polygon — IGNORE detections inside this area (ROI/RONI polygons are cleared individually inside the ROI editor)
• FPS to AI: Selector (dropdown) controlling frames sent to AI (lower = less GPU usage)
• Clear: Resets this camera's AI settings to their defaults.
• Click Save to apply changes

Notification

Channels (Delivery Methods)

Each channel can be tested, toggled active/inactive, edited, or deleted (cannot delete if rules reference it). When multiple headers are needed, separate them with a comma (,).

Rules (IF-THEN Automation)

Create rules in 5 steps. Enter a Rule Name (1-100 characters):

1. IF (Trigger): AI Event (Flame/Smoke/Falldown/Intrusion), camera status (Online/Offline), or System Alert
2. Scope & Schedule: Select cameras/groups + schedule (Always or Custom days/hours)
3. Throttle: Choose the delivery cadence — Immediate (Every Match), Once per minute, Once per hour, Batched (5 min), or Digest (Hourly) — plus an optional Max per day (optional) cap.
4. Gates (optional): Additional conditions (e.g., Confidence > 0.5, Box Size)
5. THEN (Action): Select channels and email recipients. Each rule can also choose which alert sound plays when it fires, with a preview button to hear it.

Global Policies

• Notification Countdown: Min seconds between same-event alerts from same source
• Limits: Max notifications per minute / per hour
• Quiet Hours: Suppress all notifications during specified time window

Click Save Policies to apply. Click Reset to restore defaults.

Users

Admin/Super_Admin only. Settings → Users.

• Add User: + Add → username (3-20 chars), password (4+), email, role → Save
• Edit: Change role, password, or status (email is not editable). Setting the status to Inactive prevents login without deleting the account.
• Roles: Admin, User (Super_Admin cannot be assigned in the UI)
• Role change limit: a User can be promoted to Admin, but an Admin cannot be downgraded to User.

License Management

License Status

The license page shows: current status (Active / Expiring (N days left) / Expired), expiration date, and licensed camera count.

Online Activation

1. Enter key: XXXXX-XXXXX-XXXXX-XXXXX
2. Click Activate License

Offline Activation

1. Click Generate Request File → save .req file
2. Send file to support → receive license file back
3. Click Upload License File → Apply License

System Settings

Help Page

Settings → Help (available to all roles). Contains:

• User Manual — Opens this manual in your default browser
• Contact Support — MarkAny Co., Ltd. support information (website, phone, email)
• Keyboard Shortcuts — Quick reference for all keyboard shortcuts
• Third-Party Licenses — Qt (LGPL v3), GStreamer (LGPL v2.1+), OpenCV (Apache 2.0), Rust crates (MIT/Apache 2.0)

System Maintenance

Clear cached data to free disk space (Settings → System Settings → Updates & Reports → Maintenance):

• Application Cache, Temporary Files, Thumbnail Cache
• System Logs, Video Recordings, Image Captures, Backup Storage
• Clear All Cache & Logs: One-click cleanup (confirmation required, system may be slower temporarily while rebuilding cache)

5. Incident Log

Access via the Incident Log tab in the top navigation, or right-click a server/camera → View Logs.

Filtering Incidents

The table shows: Date & Time, Thumbnail, Camera, Type, Event (True/False), Checked by, Checked at. Pagination at the bottom.

Incident Detail View

Click any row to open:

• Video playback with timeline scrubber — includes about 5 seconds of pre-event buffer. Click Play, drag to seek.
• AI bounding box overlay on the captured image showing what was detected
• Metadata: Incident Type, Server, Camera Name, Last Edit, Event T/F

Auditing Incidents

1. Select True or False — True = real event confirmed, False = false positive
2. Add notes (optional) — describe what you observed (up to 500 characters)
3. Click "Check" — the incident now records: Checked by, Last Edit, True/False status

Downloading Incidents

• Click the Download icon on the detail page
• Select format: Image (PNG/JPG) and/or Video (MP4/AVI)
• Optionally enable the Include Bounding Box checkbox to burn the AI detection boxes into the exported image or video
• Choose a save folder → download begins with progress indicator

Additional Export Options

• Generate Incident Report: Settings → System Settings → Updates & Reports → Generate Incident Report → choose date range, style, graph grouping → Generate
• Daily Report: Settings → System Settings → Updates & Reports → Download Daily Report

6. Frequently Asked Questions

7. Troubleshooting

Connection Issues

Disconnect & Reconnect Behavior

When the server connection is lost:

• A "Disconnected" overlay appears on the main screen
• Camera streams freeze on the last received frame
• The client automatically attempts to reconnect every few seconds
• Once reconnected, streams resume and the overlay disappears
• Notifications received during disconnection are not retroactively shown

Common Connection Problems

• Verify the server application is running on the target machine
• Confirm IP address and port (default: 7878)
• Check firewall allows inbound connections on the server port
• Ensure both machines are on the same network or VPN is configured
• "Server disconnected" during use → check network. Client will auto-reconnect.

Camera Issues

Feed not displaying

• Test RTSP URL in VLC first to confirm it works outside MirAI
• Ensure camera is reachable from the server machine (not just client)
• Check camera status indicator (green=online, red=offline, yellow blink=stall)
• Review server logs: logs/server.log

No cameras found during discovery

• Verify IP range is correct and cameras are powered on
• Check that cameras support ONVIF and have it enabled
• Try adding manually via RTSP URL instead

AI Detection Issues

• Confirm detection is enabled for the camera (right-click → AI Settings)
• Confirm camera is in a group with an active pipeline assigned
• Lower confidence thresholds (try 0.3-0.5 for testing)
• Check ROI/RONI — event may be outside your ROI polygon
• Verify server GPU is not overloaded (>90% → dropped frames)
• Check that the correct AI model is Active in Model Hub
• Review logs/ai_pipeline.log

Notification Issues

• Verify channel is active (toggle ON in Channels tab)
• Click Test button on channel card to confirm delivery
• Check that a rule links the event type to the channel
• Check Global Policies — Quiet Hours may be suppressing
• Check rate limits (max per minute/hour may be exhausted)
• Email: verify SMTP settings, port, encryption, credentials
• Webhook: verify URL is reachable from server

Common Error Messages

Performance Tips

• Use smaller grid layouts to reduce CPU on the client
• Switch cameras to Low Resolution in the hover menu when monitoring many feeds
• Lower the "FPS to AI" selector to reduce server GPU usage
• Set retention policy to auto-delete old recordings and free disk space
• Close other resource-intensive applications on the client machine

Quick Reference: Tips & Shortcuts

8. Glossary