- Introduced a `virtual` attribute for automatically generated section transitions and backlinks in the graph edge definitions, indicating their virtual nature.
- Updated documentation to reflect the addition of the `virtual` attribute for both section transitions and backlinks, clarifying its implications for scoring in the retriever.
- Enhanced the understanding of edge types by specifying that these automatically generated edges will receive a penalty during scoring.
- Added functionality to automatically create inverse backlinks for intra-note edges at the chunk level, ensuring that backlinks are generated only when they do not already exist.
- Updated the documentation to outline the requirements and rules for backlink creation, including conditions for deduplication and scope.
- Introduced unit tests to validate the creation of backlinks and ensure correct behavior when existing backlinks are present.
- Incremented version to 4.4.0 to reflect the new feature addition.
- Updated provenance priorities and introduced a mapping from internal provenance values to EdgeDTO-compliant literals.
- Added a new function `normalize_provenance` to standardize internal provenance strings.
- Enhanced the `_edge` function to include an `is_internal` flag and provenance normalization.
- Modified the `EdgeDTO` model to include a new `source_hint` field for detailed provenance information and an `is_internal` flag for intra-note edges.
- Reduced the provenance options in `EdgeDTO` to valid literals, improving data integrity.
Introduce a new method for persisting rejected edges for audit purposes, enhancing traceability and validation logic. Update the decision engine to utilize a generic fallback template for improved error handling during LLM validation. Revise documentation across multiple files to reflect the new versioning, context, and features related to Phase 3 validation, including automatic mirror edges and note-scope zones. This update ensures better graph integrity and validation accuracy in the ingestion process.
