Get started¶
This section takes you from nothing to a working AIGr.id setup that runs AI workloads.
Pick your starting point¶
| I want to... | Go to |
|---|---|
| See the whole journey with working examples | Quickstart tutorial |
| Start a brand-new network | 1. Create a network |
| Add my Kubernetes cluster to an existing network | 2. Join a cluster |
| Add a single machine to an existing cluster | 3. Join a node |
| Deploy models or send tasks to an existing network | Block creator guide or Inference user guide |
The setup order¶
Infrastructure is built top-down: a network must exist before clusters can join it, and a cluster must exist before nodes can join it.
flowchart LR
A["1. Create a network<br/><small>install the management cluster</small>"] --> B["2. Join a cluster<br/><small>register it and install cluster services</small>"]
B --> C["3. Join a node<br/><small>optional: add more machines</small>"]
C --> D["4. Deploy workloads<br/><small>components, blocks, vDAGs</small>"]
D --> E["5. Send tasks<br/><small>inference server or vDAG controller</small>"]
Joining someone else's network?
Skip step 1. You only need the network's Parser URL and Cluster Controller Gateway URL from its operator.
Before you begin¶
You will need:
-
The source code. Clone the repository; all commands in these docs are run from its root folder unless stated otherwise.
-
Basic Kubernetes skills. Familiarity with
kubectlandkubeadm, and access to at least one Kubernetes cluster. - Python 3 and
pip, for the onboarding scripts and SDK examples. - Docker, if you will build your own block images.
If any term is unfamiliar, keep the Glossary open in another tab.