Remove flag for automatic schema extraction

Author: NathalieCharbel

CHANGELOG.md

@@ -4,8 +4,7 @@
 
 ### Added
 
-- Added support for automatic schema extraction from text using LLMs.
-
+- Added support for automatic schema extraction from text using LLMs. In the `SimpleKGPipeline`, when the user provides no schema, the automatic schema extraction is enabled by default. 
 ## 1.7.0
 
 ### Added

docs/source/user_guide_kg_builder.rst

@@ -466,8 +466,6 @@ within the configuration file.
                 ["House", "RULES", "Planet"]
             ]
         },
-        /* Control automatic schema extraction */
-        "auto_schema_extraction": false,
         "lexical_graph_config": {
             "chunk_node_label": "TextPart"
         }
@@ -511,8 +509,6 @@ or in YAML:
         - ["Person", "PARENT_OF", "Person"]
         - ["Person", "HEIR_OF", "House"]
         - ["House", "RULES", "Planet"]
-    # Control automatic schema extraction 
-    auto_schema_extraction: false
     lexical_graph_config:
         chunk_node_label: TextPart
 

src/neo4j_graphrag/experimental/pipeline/config/template_pipeline/simple_kg_builder.py

@@ -89,7 +89,6 @@ class SimpleKGPipelineConfig(TemplatePipelineConfig):
     perform_entity_resolution: bool = True
     lexical_graph_config: Optional[LexicalGraphConfig] = None
     neo4j_database: Optional[str] = None
-    auto_schema_extraction: bool = False
 
     pdf_loader: Optional[ComponentType] = None
     kg_writer: Optional[ComponentType] = None
@@ -170,7 +169,7 @@ def _get_schema(self) -> Union[SchemaBuilder, SchemaFromTextExtractor]:
         Get the appropriate schema component based on configuration.
         Return SchemaFromTextExtractor for automatic extraction or SchemaBuilder for manual schema.
         """
-        if self.auto_schema_extraction and not self.has_user_provided_schema():
+        if not self.has_user_provided_schema():
             return SchemaFromTextExtractor(llm=self.get_default_llm())
         return SchemaBuilder()
 
@@ -234,7 +233,7 @@ def _process_schema_with_precedence(
         return entities, relations, potential_schema
 
     def _get_run_params_for_schema(self) -> dict[str, Any]:
-        if self.auto_schema_extraction and not self.has_user_provided_schema():
+        if not self.has_user_provided_schema():
             # for automatic extraction, the text parameter is needed (will flow through the pipeline connections)
             return {}
         else:
@@ -290,7 +289,7 @@ def _get_connections(self) -> list[ConnectionDefinition]:
             )
 
             # handle automatic schema extraction
-            if self.auto_schema_extraction and not self.has_user_provided_schema():
+            if not self.has_user_provided_schema():
                 connections.append(
                     ConnectionDefinition(
                         start="pdf_loader",
@@ -382,6 +381,6 @@ def get_run_params(self, user_input: dict[str, Any]) -> dict[str, Any]:
                 )
             run_params["splitter"] = {"text": text}
             # Add full text to schema component for automatic schema extraction
-            if self.auto_schema_extraction and not self.has_user_provided_schema():
+            if not self.has_user_provided_schema():
                 run_params["schema"] = {"text": text}
         return run_params

src/neo4j_graphrag/experimental/pipeline/kg_builder.py

@@ -124,9 +124,6 @@ def __init__(
                 perform_entity_resolution=perform_entity_resolution,
                 lexical_graph_config=lexical_graph_config,
                 neo4j_database=neo4j_database,
-                auto_schema_extraction=not bool(
-                    schema or entities or relations or potential_schema
-                ),
             )
         except (ValidationError, ValueError) as e:
             raise PipelineDefinitionError() from e