The first AI copilot built specifically for video management systems. Control cameras, search recordings, configure AI models, and manage your entire surveillance infrastructure through natural conversation.
Radha is not a chatbot bolted onto a dashboard. It is a purpose built AI assistant with 22 operational tools that execute real actions on your system. Powered by a 9 billion parameter model fine tuned for VMS, every conversation stays within your infrastructure.
22
Operational Tools
12
Languages Supported
9B
Parameter LLM
0
Cloud Dependencies
Most AI assistants answer questions. Radha executes operations. It connects to every part of the Visylix platform through 22 specialized tools, turning natural language into real system actions.
Talk to your VMS like you talk to a colleague. Radha understands complex requests like "show me all face detections from the parking lot cameras between 2pm and 4pm yesterday" and executes them instantly.
Radha runs a 9 billion parameter model deployed entirely within your infrastructure. No data leaves your network. Perfect for government, defense, banking, and any organization that requires complete data sovereignty.
Responses stream back instantly via Server Sent Events (SSE). You see results as they generate, not after a long wait. Tool execution happens in real time with live progress updates.
Every input passes through dedicated detection patterns before processing. Radha identifies and blocks jailbreak attempts, instruction overrides, and role impersonation within defined permissions.
Each user has individual rate limits to prevent abuse. Administrators can configure limits per role, ensuring fair usage across teams while protecting system resources.
Radha understands your system topology. It knows which cameras are in which locations, what AI models are running, and what alerts are active. Every response is informed by your actual deployment.
Radha understands the context of your surveillance system. Here are examples of how operators interact with Radha every day.
How many cameras are offline right now?
You have 3 cameras currently offline: Lobby Camera 2 (disconnected 2 hours ago), Parking Level B1 (network timeout), and Loading Dock 3 (scheduled maintenance). Would you like me to attempt reconnection on the first two?
Enable face recognition on all entrance cameras
I have enabled face recognition on 8 entrance cameras: Main Lobby, Side Entrance A, Side Entrance B, Parking Entry, Parking Exit, Loading Dock 1, VIP Entrance, and Employee Gate. All are using the default gallery with 94.2% accuracy threshold. Want me to adjust the sensitivity?
Show me crowd alerts from today with more than 50 people
Found 4 crowd detection alerts today exceeding 50 people: Cafeteria at 12:34 PM (78 people), Main Lobby at 9:02 AM (63 people), Auditorium at 2:15 PM (124 people), and Parking Level A at 5:45 PM (52 people). I can pull up the footage for any of these.
Every tool connects directly to the Visylix platform. Radha does not simulate actions. It executes them.
List all cameras with optional status filter. See which streams are online, offline, or in error state across your deployment.
Get detailed information on any specific stream including resolution, protocol, connection status, and active AI models.
Play a live camera feed directly inline. Radha opens the viewer and starts the feed in the conversation window.
Start stopped cameras or stop running streams. Requires STREAMS_EDIT permission for your user role.
Begin capturing new video from a live camera. Radha distinguishes between starting a new capture and viewing saved files.
Stop an active recording session on any camera. Radha confirms the recording duration and file size when complete.
Search and list saved recordings by date, stream, or time range. Ask "show me recordings from yesterday" and Radha handles the rest.
Play any saved recording inline by ID. Radha opens the player within the conversation with full playback controls.
Capture a JPEG snapshot from any live camera instantly. Requires SNAPSHOTS_CREATE permission.
List recent snapshots with timestamps and camera details. Filter by stream or time range.
View AI generated alerts filtered by severity, acknowledgment status, or detection type across all cameras.
Mark an alert as acknowledged. Radha confirms the action and removes it from the active alert queue.
Browse AI detections by type: face, person, vehicle, plate, object, or pose. Filter by camera and time range.
Check CPU, RAM, disk usage, and stream counts. Get a complete health overview of your infrastructure.
View total, online, and error stream counts at a glance. Radha summarizes your entire deployment status.
Check AI engine health and GPU utilization. See which models are loaded and their current inference performance.
View all recording and snapshot schedules. See frequency, target cameras, and next scheduled run time.
Create once, daily, or weekly recurring schedules for recordings and snapshots through conversation.
Remove an existing schedule by ID. Requires SCHEDULES_DELETE permission for your role.
Control video playback: pause, play, seek, mute, change speed, or go fullscreen. All through voice or text.
Navigate to any page in the Visylix dashboard. Say "go to recordings" and Radha routes you there instantly.
Radha was built for organizations where data privacy is not optional. Every component runs on premise with zero cloud dependencies.
Radha runs a 9 billion parameter model fine tuned on 1,500+ VMS training examples, deployed entirely on your hardware. No API calls to external services. No data leaves your network.
Every user input passes through dedicated detection patterns that identify jailbreak attempts, instruction overrides, and role impersonation. Radha logs and blocks manipulation attempts automatically.
Radha respects your existing permission structure. Operators can only use tools their role allows. An operator without camera deletion permissions cannot ask Radha to delete cameras.
Because Radha runs entirely on premise with no cloud dependencies, it works in completely air gapped environments. Government facilities, defense installations, and classified networks can use Radha without modification.
Radha is not a wrapper around any cloud AI service. It runs a dedicated language model on your hardware, fine tuned with over 1,500 training examples specifically for video management operations.
Every tool in Radha executes real system commands. When you ask Radha to list cameras or start a recording, it calls the Visylix API and confirms the result. This is not a simulation.
Radha can chain up to 5 rounds of tool calls in a single request. Ask it to list cameras, find one that is offline, and restart it, all in one conversation turn.
Real time streaming delivers 7 distinct event types: conversation start, content, tool calls, tool results, frontend actions, completion with suggestions, and error handling.
Context aware suggestions appear after every response. Radha recommends relevant next actions based on what you just did, so operators always know what is possible.
Schedule a live demo and experience what it feels like to manage your entire video infrastructure through conversation.