Skip to content

Converters API

Overview

Graph conversion from Pydantic models to NetworkX graphs.

Module: docling_graph.core.converters

GraphConverter

Main class for converting Pydantic models to knowledge graphs.

class GraphConverter:
    """Convert Pydantic models to NetworkX graphs."""

    def __init__(
        self,
        config: Optional[GraphConverterConfig] = None,
        node_id_registry: Optional[NodeIDRegistry] = None
    ):
        """Initialize converter."""

Methods

convert()

def convert(
    self,
    models: List[BaseModel],
    reverse_edges: bool = False
) -> nx.MultiDiGraph:
    """
    Convert Pydantic models to graph.

    Args:
        models: List of Pydantic model instances
        reverse_edges: Create reverse edges

    Returns:
        NetworkX MultiDiGraph
    """

Example:

from docling_graph.core.converters import GraphConverter

converter = GraphConverter()
graph = converter.convert(models, reverse_edges=False)

print(f"Nodes: {graph.number_of_nodes()}")
print(f"Edges: {graph.number_of_edges()}")

NodeIDRegistry

Registry for stable node ID generation.

class NodeIDRegistry:
    """Generate and track stable node IDs."""

    def get_or_create_id(
        self,
        model: BaseModel,
        node_type: str
    ) -> str:
        """
        Get or create stable ID for model.

        Args:
            model: Pydantic model instance
            node_type: Type of node

        Returns:
            Stable node ID
        """

Features: - Deterministic ID generation - Collision detection - Cross-batch consistency - Graph ID field support

Example:

from docling_graph.core.converters import NodeIDRegistry

registry = NodeIDRegistry()
node_id = registry.get_or_create_id(person_model, "Person")

GraphConverterConfig

Configuration for graph conversion.

class GraphConverterConfig(BaseModel):
    """Configuration for graph converter."""

    reverse_edges: bool = False
    include_metadata: bool = True