What is ArcisGPT and how does it work?
ArcisGPT is VMukti's generative-AI video module that lets surveillance operators query archived footage in natural language and receive ranked video clips with auto-generated summaries. Instead of scrubbing through hours of timeline, an operator types or speaks a request like "show me the silver SUV at gate 4 between 2 and 4 pm yesterday" and ArcisGPT returns matching clips, captioned events, and a written incident summary. It runs on top of VMukti Cloud VMS, works across any ONVIF camera, and sits alongside 25+ other on-platform AI models (ANPR, facial recognition, crowd, PPE, fire/smoke). ArcisGPT is part of VMukti's 26-model AI catalogue and is the foundation of the Visual Bot API that exposes the same capabilities to autonomous AI agents.
What ArcisGPT solves
Traditional video analytics fire pre-configured rules. ArcisGPT inverts the model: the operator asks a question and the AI assembles the answer from indexed video, metadata, and analytics events. This collapses investigation time from hours to seconds for the most common buyer-side workflows — post-incident review, suspect tracking, asset recovery, and SLA evidence.
How a query is processed
1. Indexing — every frame is enriched at ingest with object, attribute, scene, and event tags by the on-platform AI models. 2. Query parsing — the operator's natural-language request is parsed into a structured spec: subject (silver SUV), location (gate 4), time window (yesterday 2-4pm), and constraint (ranked by relevance). 3. Retrieval — the spec is matched against the index across one or many cameras. 4. Generation — top-ranked clips are returned with auto-written captions and a one-paragraph incident summary the operator can paste into a report.
Where it fits in the platform
- VMS layer — ArcisGPT consumes the same video archive as the standard VMukti VMS timeline; no separate storage tier.
- AI layer — sits alongside ANPR, face recognition, crowd, PPE, intrusion, fire/smoke, and 20 other on-platform models.
- ICCC layer — operators in an Integrated Command Control Centre can issue ArcisGPT queries from the main operator UI and surface results on the video wall.
- Visual Bot API — exposes ArcisGPT as a programmable agent so customers can schedule recurring inspections ("every hour, flag any PPE violation at the loading bay").
Where it does not fit
ArcisGPT is not a replacement for real-time alerting — for sub-second perimeter detection or ANPR-driven gate control, the rule-based and CV models still own the path. ArcisGPT shines for post-event search and multi-camera correlation, where natural-language framing dramatically reduces operator workload.
Procurement notes
ArcisGPT is available as an add-on module to VMukti Cloud VMS for both commercial and 889-safe (US federal) deployments. There is no separate hardware requirement; the module runs inside the same cloud control plane as the rest of the platform.
Last reviewed: 2026-05-20