What is the AI Assistant?
The AI Assistant in Cordatus is a conversational control layer over the platform. Devices, cameras, AI jobs, alarms, users and roles — much of what you do in the Cordatus dashboard can be done in plain language. Instead of hunting through menus, you simply describe what you want, and the assistant turns your intent into real, permission-scoped platform actions.
It sits on top of the platform you already trust — it does not replace it. Every action the assistant performs is the same action available in the dashboard, executed through the same permission-scoped APIs, so results are always real, scoped, and auditable.
Why the AI Assistant?
Unlike a generic chatbot that only answers questions, the Cordatus AI Assistant executes real operations across the platform:
- Natural language: No commands or syntax to memorize — describe what you need in your own words.
- Real actions: Calls the same permission-scoped platform APIs used by the dashboard. Every action is real and auditable.
- Permission-scoped: The assistant can only do what your account is allowed to do — it inherits your exact roles and permissions.
- Multilingual: Understands and replies in your language.
- Structured answers: Results come back as tables and summaries built from live platform data, not free text.
- Asks instead of guessing: When a request could match more than one thing, the assistant lists the candidates and asks you to choose.
- Context-aware: Multi-turn flows keep their context, so short replies like "yes" or "the second one" land in the right place.
Who is it for?
Operators:
- No menu-hunting — describe what you need and the platform does it.
- Fast, everyday lookups like checking a device, finding a camera, or reviewing alarms.
Engineers & Administrators:
- The assistant calls the same permission-scoped platform APIs you already use — every action is real, scoped, and auditable.
- Manage devices, cameras, jobs, alarms, users and roles without leaving the conversation.
Multi-site teams:
- Check devices and run AI jobs across sites from a single conversational interface.
Core Capabilities
| Domain | What you can do |
|---|---|
| Devices & Monitoring | List devices, connect to them, read live CPU / GPU / memory / temperature and analytics |
| Camera Management | List IP and physical cameras and their details (it cannot open a live stream) |
| AI Jobs, Engines & Models | List and start jobs, bulk delete them, list available engines and models |
| Alarms & Notifications | Create, list, update and delete device and inference alarms; read triggered events; list channels |
| Users, Roles & Organization | List organizations and members, manage roles and asset permissions |
Key Concepts
- AI Assistant — The conversational interface that turns natural language into real platform actions.
- Intent — What you want to happen, expressed in your own words.
- Action / Command — The real platform operation the assistant performs on your behalf (e.g., starting a job, creating an alarm).
- Selection card — An interactive card shown when a request needs clarification (e.g., which camera or device).
- Confirmation — Before some destructive operations (for example deleting an alarm) the assistant asks you to confirm. This is not guaranteed for every state-changing action — see the warning in the Quickstart.
- Permission scope — The set of roles and permissions on your account that bound everything the assistant can do.
How it works
- You describe what you want in plain language (any supported language).
- The assistant understands your intent and routes it to the right domain (devices, cameras, jobs, alarms, users or roles).
- If anything is ambiguous, it lists the candidates (which camera? which device?) instead of guessing.
- It performs the real action through the same permission-scoped APIs as the dashboard.
- You see the result — a table of devices, a started job, or a summary of what changed.
New to the assistant? Start with the Quickstart to run your first command in a couple of minutes.