graphrag and youtu-graphrag

Microsoft GraphRAG is a general-purpose graph-based RAG framework, while Youtu-GraphRAG is a specialized research implementation focusing on unified agentic reasoning over graphs—making them **complements** that could be combined, though they target different use cases (production systems vs. research-driven complex reasoning).

graphrag
76
Verified
youtu-graphrag
51
Established
Maintenance 20/25
Adoption 11/25
Maturity 25/25
Community 20/25
Maintenance 10/25
Adoption 10/25
Maturity 9/25
Community 22/25
Stars: 31,429
Forks: 3,319
Downloads:
Commits (30d): 7
Language: Python
License: MIT
Stars: 1,082
Forks: 161
Downloads:
Commits (30d): 0
Language: Python
License:
No risk flags
No Package No Dependents

About graphrag

microsoft/graphrag

A modular graph-based Retrieval-Augmented Generation (RAG) system

Extracts knowledge graphs from unstructured text using LLMs, then uses those graph structures to improve retrieval and reasoning for private data. Implements a data indexing pipeline that transforms narrative documents into entity-relationship graphs, enabling more contextual and discovery-oriented query responses compared to standard vector retrieval. Supports prompt tuning workflows and integrates with major LLM providers through a configuration-driven architecture.

About youtu-graphrag

TencentCloudADP/youtu-graphrag

[ICLR 2026] Youtu-GraphRAG: Vertically Unified Agents for Graph Retrieval-Augmented Complex Reasoning

Implements a unified agentic framework with schema-guided hierarchical knowledge graph construction across four levels (attributes, relations, keywords, communities) and a novel "dually-perceived" community detection algorithm that fuses structural topology with semantic information. Core capabilities include schema-aware query decomposition with iterative reflection (IRCoT), FAISS-accelerated retrieval, and a four-tier architecture supporting domain adaptation through minimal schema intervention, all configurable via centralized YAML management.

Scores updated daily from GitHub, PyPI, and npm data. How scores work