Artica AI
The assistant of the Artica Web console: principles, operating modes, language model choices and the pages of this section.
Artica AI is the assistant of the Artica Web console.
It reads the state of the appliance, explains what it sees and proposes actions.
It runs on the appliance itself, as a dedicated service.
The AI proposes, the policy decides
The language model never decides anything on its own.
It writes the chat answers, explains incidents and suggests diagnostics.
Every action goes through a deterministic policy, then through a transaction that is verified and can be rolled back.
Artica AI never runs a free shell command.
It only uses declared actions, each with a risk level from L0 (read only) to L5.
Artica: AI harness/agentic
- Connecting an Artica appliance to DeepSeek:
Connect the assistant to the hosted DeepSeek API in a few minutes.
- Connecting an Artica appliance to the Artica AI appliance
Use the AI appliance of your network as the model of your other appliances.
Connect the Artica AI assistant of an appliance to an Artica AI appliance of your network
Artica AI Appliance:
- Artica AI appliance:
Concepts, prerequisites, performance and use cases.
- Installing an Artica AI appliance:
Step-by-step installation and connection of the client appliances.
- Explaining and optimizing the Artica AI appliance:
where the time goes and how to tune it.
Operating modes
| Mode | What Artica AI may do |
|---|---|
| Advisor | Read and explain only. No change is made. |
| Copilot | Read freely. Every change waits for the approval of an administrator. |
| Autopilot | Run a short list of low-risk actions by itself, inside time windows and an hourly limit. Everything else still waits for approval. |
A safe mode and a kill switch stop every change at once.
Each decision and each action is written to an audit log.
The language model
Artica AI talks to any OpenAI-compatible server.
Three choices are possible:
- A hosted API, on the Internet.
- A model server of your own, such as Ollama or vLLM.
- An Artica AI appliance: an Artica appliance that serves a local model to the others.
Without any model, Artica AI keeps working in deterministic mode.
It then answers with its own diagnostics, without free text written by a model.