Description

This PR adds an optional schema enforcement layer to validate the extracted entities, relations, and properties against the schema that is passed to the LLM as guidance for the extraction. This change introduces:

• A new _clean_graph function in the LLMEntityRelationExtractor class that filters out invalid nodes, relationships, and properties based on a provided SchemaConfig object:
• Support for different enforcement modes (e.g., NONE vs. STRICT). So far implemented using Enum.
• Unit tests covering various enforcement scenarios (no schema, invalid nodes, invalid relations, etc.).

The cleanup is performed at the chunk graphs level after the extraction and before post-processing, i.e.,before relationships to the chunks are created.

Cleanup logic:

• Remove invalid nodes (not conformant to the schema)
• Remove invalid node properties (not conformant to the schema)
• Remove nodes that are left with no properties
• Remove invalid relations (whose types or start/end nodes are not conformant to the schema)
• Remove relations whose start/end node was removed
• Remove invalid relations properties (not conformant to the schema).

Type of Change

• New feature
• Bug fix
• Breaking change
• Documentation update
• Project configuration change

Complexity

Note

Please provide an estimated complexity of this PR of either Low, Medium or High

Complexity:

How Has This Been Tested?

• Unit tests
• E2E tests
• Manual tests

Checklist

The following requirements should have been met (depending on the changes in the branch):

• Documentation has been updated
• Unit tests have been updated
• E2E tests have been updated
• Examples have been updated
• New files have copyright header
• CLA (https://neo4j.com/developer/cla/) has been signed
• CHANGELOG.md updated if appropriate