Graph

Core

OrkesGraph(state[, name, description, traced])

A class to represent a stateful graph for orchestrating multi-agent workflows.

Runner

GraphRunner(graph_name, graph_description, ...)

Executes a compiled OrkesGraph, managing state and tracing the execution.

Schema

NodePoolItem	Represents an item in the node pool, which is a collection of nodes and their associated edges within the graph.
NodeTrace	Represents the trace of a single node's execution within the graph.
LLMTraceSchema	Represents the input and output of an LLM interaction for tracking purposes.
FunctionTraceSchema	Represents the trace of a single function's execution, capturing its inputs, output, and timing.
EdgeTrace	Represents the trace of a single edge traversal during a graph execution.
TracesSchema	Represents the complete execution trace of a graph run.

Units

Node(name, func, graph_state)	Represents a node in the computational graph.
Edge(from_node[, to_node, max_passes])	Represents a directed connection between two nodes in the graph.
ForwardEdge(from_node, to_node[, max_passes])	An edge that unconditionally forwards data and control from a source node to a destination node.
ConditionalEdge(from_node, gate_function, ...)	An edge that determines the next node based on the outcome of a gate function.
ParallelEdge(from_node, to_nodes, ...[, ...])	An edge that creates multiple parallel branches of execution.

Utilities

orkes_tracable(func)	A decorator that traces the input, output, and execution time of a function and records it in the current edge's trace.
function_assertion(func, expected_type)	Asserts that a function has at least one parameter with the expected type annotation.
is_typeddict_class(obj)	Checks if an object is a TypedDict class.
check_dict_values_type(d, cls)	Checks if all values in a dictionary are of a certain type.
randomize_color_hex()	Generates a random hex color code.