Skip to main content

Nodes, Edges, and Walkers in JIVAS

JIVAS leverages Nodes, Edges, and Walkers from JacLang to build and traverse graph-based architectures. These primitives form the foundation of the framework's modular and extensible design.

Nodes​

Nodes are the building blocks of the graph. They represent entities with properties and behaviors, similar to classes in object-oriented programming.

Key Features of Nodes​

  • Attributes: Nodes can have attributes (e.g., has name: str).
  • Abilities: Nodes can define callable methods (can) and visit-dependent methods triggered by walkers.

Example​

node ExampleNode {
    has name: str = "Default Name";

    can greet {
        print(f"Hello, my name is {self.name}!");
    }
}

Connecting Nodes​

Nodes can be connected to form relationships:

  • One-to-One: A single node connects to another.
  • One-to-Many: A single node connects to multiple nodes.
  • Many-to-One: Multiple nodes connect to a single node.
  • Many-to-Many: Groups of nodes connect to each other.
node NodeA {}
node NodeB {}

with entry {
    node_a = NodeA();
    node_b = NodeB();
    node_a ++> node_b;  # One-to-One connection
}

Edges​

Edges define relationships between nodes. They can be generic or custom, with properties.

Generic Edges​

with entry {
    node_1 ++> node_2;  # Unidirectional edge
    node_1 <++> node_2; # Bidirectional edge
}

Custom Edges​

Custom edges allow additional properties:

edge CustomEdge {
    has weight: int;
}

with entry {
    node_1 +:CustomEdge:weight=5:+> node_2;
}

Deleting Edges​

Edges can be removed using the del keyword:

node_1 del --> node_2;

Walkers​

Walkers are agents that traverse the graph, performing tasks on nodes they visit. They are essential for executing workflows in JIVAS.

Key Features of Walkers​

  • Attributes: Walkers can have their own properties.
  • Abilities: Walkers define callable and visit-dependent methods.
  • Graph Traversal: Walkers navigate the graph using the visit keyword.

Example​

walker ExampleWalker {
    can on_node with ExampleNode entry {
        print(f"Visiting node: {here.name}");
    }
}

Spawning Walkers​

Walkers are spawned at specific points in the graph:

spawn ExampleWalker() on node_a;

Traversal​

Walkers use visit to navigate:

  • visit [-->]: Visit successors.
  • visit [<--]: Visit predecessors.
  • visit [node_name]: Visit a specific node.

Summary​

  • Nodes: Represent entities with attributes and abilities.
  • Edges: Define relationships between nodes.
  • Walkers: Traverse the graph, executing tasks on nodes.

By combining these primitives, JIVAS enables developers to build intelligent, graph-based AI solutions.