What runs where

Concept 2 · 3:42

The video shows where each part of Weik.io runs, from the weikio CLI and the admin UI on your machine to your own agents and the Weik.io installation they link to. Weik.io places a deployment on every agent that matches its requirements.

Download the mp4Recorded on 2026.2

What it covers

The weikio CLI on your machine, the admin UI in a browser, and a hub that runs no integrations
Agents with a leaf node that connects outward to the hub
Agent capabilities and allowed projects, set in the admin UI under Settings › Agents
Placement of sandbox/hello by the requirements in integration.yaml
One copy per matching agent, one on each host
A refused deployment when no online agent meets a requirement

Commands in this video

This video explains with drawings and types no commands.

Transcript

0:01An integration runs somewhere. This video shows what runs it.
0:09What runs an integration is an agent, a program on your own host.
0:14The Weik.io CLI and the admin UI talk to your Weik.io installation.
0:20The hub decides where each deployment runs but runs no integration itself.
0:25Agent dev on the hub's host carries devtest only,
0:29so production deployments never land on it.
0:34The drawing marks each agent host as a site.
0:41This host, marked site fi-01, holds agent fi-01 and its leaf node.
0:49The leaf node is a small relay that keeps one
0:52link out to the address your installation answers at.
0:56Nothing connects in, so the host needs no inbound port.
1:00The agent starts a Java process, a JVM, for each integration it runs.
1:09An administrator writes capabilities on an agent as key-value pairs,
1:14at enrolment or later.
1:16Agent fi-01 carries os=linux, region=fi and environment=production.
1:27When the agent connects, it reports its operating system and architecture.
1:31Those reports never enter the capability list,
1:34so a deployment matches only what an administrator wrote.
1:38Environment is a capability like the others,
1:41so this pair sets which environments the agent may run.
1:48An administrator writes a second list on the agent: the projects it may run.
1:52Agent fi-01 may run the projects sandbox and finance.
1:58Both lists are written in the admin UI, never by the agent.
2:03A deployment from any other project never runs on this agent.
2:11The integration sandbox/hello states its requirement, os linux, in integration.yaml.
2:18The deployment adds the environment it goes to,
2:21so the check list holds three items, the project among them.
2:28Placement checks every online agent against that list.
2:32Agent fi-01 carries all three, so the deployment runs there.
2:41Agent fi-02 carries the same capabilities and the same two projects.
2:47Placement takes every agent that matches, so the integration now runs on both.
2:56The file gains a second requirement, region se.
3:00No online agent carries region se, so the deployment is refused.
3:06The refusal names what no agent has and says
3:09what an administrator must write to fix it.
3:16The agent is what runs an integration, on a host your organization owns.
3:22Capabilities and projects come from the administrator, not from the agent.
3:27The file and the deployment's environment give the requirements,
3:31and placement takes every agent that matches.

Next in this track

Concept 3 · 3:23 Which part do I need?