mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-05-25 18:36:59 +08:00
62a1333cf2
… ### What problem does this PR solve? Closes #13857 Parent-child chunking was introduced in v0.23.0 but is only configurable through the web UI. Users managing datasets programmatically cannot enable it via the HTTP API or Python SDK because `ParserConfig` uses `extra="forbid"`, rejecting the `children_delimiter` field at validation. ### What does this PR change? Adds a `parent_child` nested config to `ParserConfig`, following the same pattern as `raptor` and `graphrag`: ```json "parser_config": { "parent_child": { "use_parent_child": true, "children_delimiter": "\n" } } ``` - api/utils/validation_utils.py — new ParentChildConfig model, added to ParserConfig - api/utils/api_utils.py — naive defaults + flatten to children_delimiter for the execution layer - api/apps/services/dataset_api_service.py — flatten on the update path - test/testcases/configs.py — updated DEFAULT_PARSER_CONFIG - test/testcases/test_http_api/test_dataset_management/test_create_dataset.py — 4 valid + 2 invalid test cases No changes to the execution layer (rag/app/naive.py, rag/nlp/search.py). Existing UI flow via ext is unaffected. ### Type of change - [ ] Bug Fix (non-breaking change which fixes an issue) - [x] New Feature (non-breaking change which adds functionality) - [ ] Documentation Update - [ ] Refactoring - [ ] Performance Improvement - [ ] Other (please describe): <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **New Features** * Added parent-child chunking configuration for dataset creation and updates with new `use_parent_child` toggle and customizable `children_delimiter` setting to specify how parent chunks are split into child chunks. * **Documentation** * Updated HTTP and Python API references with parent-child chunking configuration details and examples. <!-- end of auto-generated comment: release notes by coderabbit.ai -->