​
Usage

from morphik import Morphik

db = Morphik()

# Update a graph with additional documents
updated_graph = db.update_graph(
    name="research_graph",
    additional_filters={"category": "new_research"},
    additional_documents=["doc_123", "doc_456"]
)

print(f"Graph now has {len(updated_graph.entities)} entities")
print(f"Graph now has {len(updated_graph.relationships)} relationships")

​
Parameters

  • name (str): Name of the graph to update
  • additional_filters (Dict[str, Any], optional): Optional additional metadata filters to determine which new documents to include
  • additional_documents (List[str], optional): Optional list of additional document IDs to include
  • prompt_overrides (GraphPromptOverrides | Dict[str, Any], optional): Optional customizations for entity extraction and resolution prompts

​
Returns

A Graph object representing the updated knowledge graph.

​
Description

This method processes additional documents matching the original or new filters, extracts entities and relationships, and updates the graph with new information.

The graph update operation:

  1. Retrieves additional documents based on filters and/or specific document IDs
  2. Extracts entities and relationships from these documents
  3. Intelligently merges new entities and relationships with the existing graph
  4. Returns the updated graph with all entities and relationships

​
Advanced Examples

​
With Entity Resolution Examples

from morphik.models import EntityResolutionPromptOverride, EntityResolutionExample, GraphPromptOverrides

# Update with custom entity resolution examples
updated_graph = db.update_graph(
    name="research_graph",
    additional_documents=["doc_123"],
    prompt_overrides=GraphPromptOverrides(
        entity_resolution=EntityResolutionPromptOverride(
            examples=[
                EntityResolutionExample(
                    canonical="Machine Learning", 
                    variants=["ML", "machine learning", "AI/ML"]
                ),
                EntityResolutionExample(
                    canonical="Natural Language Processing", 
                    variants=["NLP", "natural language processing", "text processing"]
                )
            ]
        )
    )
)

# With custom entity resolution prompt template
updated_graph = db.update_graph(
    name="research_graph",
    additional_filters={"year": "2025"},
    prompt_overrides=GraphPromptOverrides(
        entity_resolution=EntityResolutionPromptOverride(
            prompt_template="""I have extracted the following entities from the text:
            
{entities_str}

Here are examples of how different entity references can be grouped together:

{examples_json}

Please resolve these entities by identifying which mentions refer to the same entity.
Group them together, selecting a canonical/preferred form for each group.
Return your resolution in JSON format with the canonical form and all its variants.
""",
            examples=[
                EntityResolutionExample(
                    canonical="General AI", 
                    variants=["AGI", "General Artificial Intelligence", "General AI"]
                )
            ]
        )
    )
)

​
With Entity Extraction Examples

from morphik.models import EntityExtractionPromptOverride, EntityExtractionExample, GraphPromptOverrides

# Update with custom entity extraction examples
updated_graph = db.update_graph(
    name="medical_graph",
    additional_filters={"category": "new_medical_data"},
    prompt_overrides=GraphPromptOverrides(
        entity_extraction=EntityExtractionPromptOverride(
            examples=[
                EntityExtractionExample(label="Insulin", type="MEDICATION"),
                EntityExtractionExample(label="Diabetes", type="CONDITION"),
                EntityExtractionExample(label="Heart rate", type="VITAL_SIGN"),
                EntityExtractionExample(label="Cardiology", type="SPECIALTY")
            ]
        )
    )
)

# With custom entity extraction template
updated_graph = db.update_graph(
    name="legal_graph",
    additional_documents=["contract1", "contract2"],
    prompt_overrides=GraphPromptOverrides(
        entity_extraction=EntityExtractionPromptOverride(
            prompt_template="""Extract legal entities from the following document:

{content}

Focus on these types of entities:
{examples}

Return the extracted entities in JSON format with the following structure:
[
{"label": "entity name", "type": "ENTITY_TYPE", "properties": {"key": "value"}}
]
""",
            examples=[
                EntityExtractionExample(
                    label="John Smith",
                    type="PERSON",
                    properties={"role": "Plaintiff"}
                ),
                EntityExtractionExample(
                    label="Acme Corporation",
                    type="ORGANIZATION",
                    properties={"type": "Corporation"}
                ),
                EntityExtractionExample(
                    label="January 15, 2025",
                    type="DATE"
                )
            ]
        )
    )
)

​
Notes

  • The graph name must match an existing graph that the user has access to.
  • Either additional_filters or additional_documents (or both) should be provided; otherwise, no new content will be added to the graph.
  • When using additional_filters, these are applied in addition to any filters used during graph creation.
  • The prompt_overrides are applied only to this update operation and do not permanently change the configuration of the graph.

Was this page helpful?

Suggest editsRaise issue