v3.0.2
Breaking Changes📦 graphragView on GitHub →
⚠ 1 breaking✨ 6 features🐛 2 fixes🔧 8 symbols
Summary
This patch introduces significant enhancements to data handling with new TableProvider abstractions, CSV support, and asynchronous reading capabilities. It also removes the NetworkX dependency from graph utilities.
⚠️ Breaking Changes
- The NetworkX dependency has been removed from graph utilities. Code relying on NetworkX for graph operations in graphrag.graphs must be updated to use the new DataFrame-based implementations.
Migration Steps
- If using graph utilities, update code to use DataFrame-based implementations instead of NetworkX in graphrag.graphs.
- Document ID, human_readable_id, and raw_data initialization has moved from create_final_documents into load_input_documents and load_update_documents; review these functions if custom initialization logic was present.
✨ New Features
- Added CSVTableProvider for table storage.
- Introduced DataReader class for typed dataframe loading from TableProvider in indexing workflows and query CLI.
- Added TableProvider abstraction for table-based storage operations.
- Added async iterator support to InputReader, utilized in load_input_documents and load_update_documents workflows.
- Added table provider factory.
- Added profiling to measure memory usage.
🐛 Bug Fixes
- Fixed missed compatibility issue related to Python 3.13.
- Removed unnecessary response format check (Fixes: #2203).