The problem
I build big topologies, often 50 or more nodes. I also use an AI agent to help me configure and check the network. The agent talks to the GNS3 server API.
The problem: when the topology is big, the AI agent must load the whole topology before it can work. All nodes, all links. This is too much. The agent gets slow, it costs more, and it loses focus. It is like asking someone to read a full map every time they only need one street.
As a network engineer, I already think in parts: "this is the core", "this is the access layer", "this is branch site A". But the GNS3 topology has no way to save these parts. The only thing close is a drawing (a rectangle), and that is just a picture. It does not tell the server which nodes are inside it.
What I want
A way to save a group of nodes as a zone. Then the AI agent (or I) can ask: "give me only zone A". The server returns only the nodes and links inside zone A, plus the links that connect zone A to the rest. So the agent works on one part at a time.
Example: 50 nodes, split into 10 zones. I can run 10 small agents, one per zone. Each agent loads only its own zone. Fast, cheap, and clear.
Suggested API
-
Create a zone: name, description, color, and a list of node ids.
POST /projects/{project_id}/zones
-
List / show / update / delete a zone.
-
A node can be in more than one zone (for example, both "edge" and "site-A").
-
The important one: get a zone's sub-topology.
GET /projects/{project_id}/zones/{zone_id}/topology
It returns:
- the nodes in the zone,
- the links inside the zone,
- the links that go out of the zone, with the node on the other side.
-
Save zones in the .gns3 file, so they are not lost when you close and open the project.
Questions
- Can a node be in many zones? I think yes.
- Only nodes in a zone, or also drawings? I think nodes only.
- When a link goes between two zones, show it in both? I think yes.
- Name: zone, domain, or region? I like "zone", but any name is fine.
Not in this request
- This is not about user permissions or access control. It is only about grouping nodes.
- This does not replace drawings. Drawings are for the eyes. Zones are for the data.
The problem
I build big topologies, often 50 or more nodes. I also use an AI agent to help me configure and check the network. The agent talks to the GNS3 server API.
The problem: when the topology is big, the AI agent must load the whole topology before it can work. All nodes, all links. This is too much. The agent gets slow, it costs more, and it loses focus. It is like asking someone to read a full map every time they only need one street.
As a network engineer, I already think in parts: "this is the core", "this is the access layer", "this is branch site A". But the GNS3 topology has no way to save these parts. The only thing close is a drawing (a rectangle), and that is just a picture. It does not tell the server which nodes are inside it.
What I want
A way to save a group of nodes as a zone. Then the AI agent (or I) can ask: "give me only zone A". The server returns only the nodes and links inside zone A, plus the links that connect zone A to the rest. So the agent works on one part at a time.
Example: 50 nodes, split into 10 zones. I can run 10 small agents, one per zone. Each agent loads only its own zone. Fast, cheap, and clear.
Suggested API
Create a zone: name, description, color, and a list of node ids.
POST /projects/{project_id}/zonesList / show / update / delete a zone.
A node can be in more than one zone (for example, both "edge" and "site-A").
The important one: get a zone's sub-topology.
GET /projects/{project_id}/zones/{zone_id}/topologyIt returns:
Save zones in the
.gns3file, so they are not lost when you close and open the project.Questions
Not in this request