Features

Everything below ships in the same package and works on the free tier — licensing only changes how many cameras you can run.

Cameras & live video

Camera onboarding: ONVIF WS-Discovery, vendor-native discovery, and active subnet scanning for cameras on routed networks. Credentials live in an encrypted vault; organize with groups, tags, and profiles.
Native vendor drivers: Cameras run native integrations per capability — Hikvision, Dahua, Axis, Hanwha, Uniview, Vivotek — and fall back to generic ONVIF/RTSP everywhere else. The UI badges each camera Native or Generic.
Sub-second live view: WebRTC live video with LL-HLS fallback, a multi-camera grid, and on-demand H.264 transcode for HEVC/MPEG-4 sources.
PTZ control: Continuous pan/tilt/zoom and preset recall for cameras that support it, native or ONVIF.
Health watchdog: Every camera is probed continuously; offline cameras are flagged in under 30 seconds, with a health dashboard covering cameras, storage, and services.
Device settings: Read and change camera time/NTP, image, encoding, and OSD settings from inside CoreSight — no more visiting each camera’s own web page.

Recording modes: Continuous, scheduled, or motion-triggered with pre/post-roll — configured per camera, recorded as crash-safe fMP4 segments.
Storage pools: Local disks and SMB/NFS/iSCSI network storage, with balancing across pools, optional mirrored redundancy, and retention by age, size, or capacity. Legal hold exempts footage from cleanup.
Timeline playback: A day timeline per camera with recording coverage at a glance, synchronized multi-camera playback, 0.5–8× speed, and frame stepping.
Export & evidence: Cut clips straight from the timeline, or build evidence bundles across cameras and time ranges for handover.

Object detection & tracking: YOLO-based detection with object tracking, line-cross and zone counting, and activity heat maps — all computed on your own hardware.
Face recognition: Detect faces, match against your identity list, and search by person.
License plate recognition: ALPR with Arabic-to-Latin normalization, feeding search and access-control rules.
Fire & smoke detection: Early-warning detection running on the same pipeline.
Smart search: Search detections by class, camera, and time — and jump straight into the recording at that moment.
GPU acceleration: CUDA, TensorRT, and OpenVINO execution providers with automatic CPU fallback.

Alarm lifecycle: Motion and analytics events coalesce into alarms with acknowledge, assign, and clear workflows plus a full history.
Escalation & notifications: Escalation policies notify the right people via webhook, email, or per-device web push.
Search & export: Query events and alarms, export CSV, and bundle related clips.
Access control: Identity and plate allowlists can drive HTTP door controllers, default-deny.
Maps: Upload floor plans and pin cameras with live status and alarm pulses.

Role-based access: Three built-in roles plus custom roles over a fine-grained permission catalog, scopable down to individual cameras.
Tamper-evident audit: A hash-chained, append-only audit log records every consequential action and verifies its own integrity from the UI.
Hardened deployment: Non-root systemd services with syscall filtering, default-deny firewall, automatic security updates, and a backend that only listens on loopback behind nginx TLS.
In-app updates: One button checks the release feed, verifies checksums, and applies the update — air-gapped boxes update from an offline bundle.
Offline licensing: Perpetual per-camera licenses verified with public-key cryptography — no phone-home, and the free tier (4 cameras) always works.

Want the operational detail? The guide book covers every feature step by step.