RAG en kennissystemen

graphrag

microsoft/graphrag

A graph-based text indexing and question-answering research project, now largely in maintenance mode.

★ 35,9KSterren
⑂ 3,8KForks
45Openstaande problemen
PythonTaal
MITLicentie
Q90Redactionele score

Overzicht

GraphRAG explores whether entity relationships and community context help answer questions across a corpus. It is useful to evaluate for broad thematic questions, alongside conventional chunk retrieval. The current README says new features and PRs are no longer accepted; fixes and dependency updates may continue. It is not an officially supported Microsoft product.

Belangrijkste functies

  • Graph indexing
  • Community context
  • Entity-oriented retrieval
  • Prompt tuning

Vereisten, installatie en snel aan de slag

Follow the command-line quickstart with a small corpus and configured model service. Back up prompts and configuration before version migration.

Gebruik

Compare local factual queries and corpus-wide questions against a chunk-retrieval baseline. Track evidence coverage, entity errors, and indexing costs.

How it works
A pipeline extracts structure from text and prepares graph-organized context and summaries for querying. Domain prompts affect index quality and indexing incurs model cost.

Audience and requirements
Graph retrieval researchers and teams able to maintain the stack. Model provider, text corpus, index storage, and indexing budget.

Practical use cases
Cross-document themes; entity relationship exploration; graph retrieval research.

Limitations and selection
Maintenance mode, indexing expense, and extraction errors are material adoption constraints. Relationships do not establish causality.

Related projects and selection
qdrant/qdrant:Baseline: compare graph indexing with vector retrieval before accepting added cost.

google/langextract:Possible complement: inspect entity evidence with extraction; custom data conversion is required.

Source review
Editorial analysis of upstream sources, without runtime or benchmark testing. Proposed workflows are editorial suggestions.

Modelcompatibiliteit en gebruiksscenario's

Configure models for extraction, summarization, and querying; validate provider and version support.

Licentie- en risiconotities

The repository page identifies MIT. Read LICENSE; model weights and datasets may have separate terms.

Editorial source review 2026-09-09T05:00:00.950Z. README and live repository page verified; current stars/forks from GitHub HTML. Last-push metadata retained from 2026-09-05 discovery snapshot. No runtime benchmark. Integration proposals are editorial analysis.

Release en onderhoud

Reviewed 2026-09-09. Counters come from repository pages; features are based on upstream documentation. See Releases in the source links. Editorial analysis of upstream sources, without runtime or benchmark testing. Proposed workflows are editorial suggestions.

milvus

milvus-io/milvus

★ 45,7KGo

LanceDB

lancedb/lancedb

★ 11,1KRust

EvalScope

modelscope/evalscope

★ 3,2KPython

llm-app

pathwaycom/llm-app

★ 58,9KJupyter Notebook

onyx

onyx-dot-app/onyx

★ 32KPython

FastGPT

labring/FastGPT

★ 29,6KTypeScript