●Change8

v2.0.0

Breaking Changes
📦 graphragView on GitHub →
⚠ 5 breaking✨ 11 features🐛 7 fixes🔧 9 symbols

Summary

Version 2.0.0 introduces major architectural changes, including reworking the API to accept callbacks, reorganizing workflows, and refactoring configuration management into type-safe dataclasses. This release also adds features like LLM provider registration and multi-index querying support.

⚠️ Breaking Changes

  • Rework API to accept callbacks: Existing API calls that relied on the old structure may need updates to pass callbacks.
  • Remove config inheritance, hydration, and automatic env var overlays: Configuration loading is now stricter; users must explicitly define configurations instead of relying on implicit inheritance or environment variable overlays.
  • Rework the update output storage structure: Code relying on the previous structure for storing workflow outputs will need updates.
  • Refactor config defaults from constants to type-safe, hierarchical dataclass: Configuration access patterns might have changed due to the move from constants to dataclasses.
  • Remove unused columns and change property document_attribute_columns to metadata: Any code referencing the old property name 'document_attribute_columns' must be updated to use 'metadata'.

Migration Steps

  1. Review API usage for any changes related to accepting callbacks.
  2. Update configuration loading logic to explicitly define settings, as automatic inheritance and environment variable overlays have been removed.
  3. Update code referencing output storage structures based on the rework.
  4. Update code referencing configuration defaults to use the new dataclass structure.
  5. Replace usage of 'document_attribute_columns' with 'metadata' in relevant configuration or access points.

✨ New Features

  • Add children to communities to avoid re-compute.
  • Add LLM Manager and Factory, to support provider registration
  • Add NLP graph extraction.
  • Add pipeline_start and pipeline_end callbacks.
  • Add caching to NLP extractor.
  • Add vector store id reference to embeddings config.
  • Export NLP community reports prompt.
  • Add dynamic retry logic.
  • Add option to prepend metadata into chunks
  • Implemented multi-index querying for api layer
  • Multi index query cli support

🐛 Bug Fixes

  • Fix DRIFT search on Azure AI Search.
  • Fix StopAsyncIteration catch.
  • Fix missing embeddings workflow in FastGraphRAG.
  • Fix proper use of n_depth for drift search
  • Fix report generation recursion.
  • Fix summarization over large datasets for inc indexing. Fix relationship summarization
  • Patch json mode for community reports

🔧 Affected Symbols

APILLM ManagerLLM Factoryworkflow runnerNLP extractorFastGraphRAGconfig defaultsquery modulefnllm