Retrieval, augmentation and generation, plus the graph approaches that connect facts rather than merely matching them. · ranked by fitness for retrieval work, then adoption · verified 13 Aug 2026

#	Name	Kind	Graph	Language	Licence	GitHub stars	Status
1	LlamaIndex	Framework	Yes, property graph index	Python, TypeScript	MIT	~45k	current
2	LangChain & LangGraph LangChain	Framework	Via integrations	Python, TypeScript	MIT	~120k	current
3	Haystack deepset	Framework	Via integrations	Python	Apache 2.0	~20k	current
4	RAGFlow InfiniFlow	Platform	Yes	Python	Apache 2.0	~35k	current
5	GraphRAG Microsoft	Library	Yes, core	Python	MIT	~25k	current
6	HLightRAG HKU Data Science	Library	Yes, core	Python	MIT	~18k	current
7	Graphiti Zep	Library	Yes, temporal	Python	Apache 2.0	~12k	current
8	Neo4j	Database	Yes, native	Cypher	GPL-3.0 / commercial	~14k	current
9	Cognee	Framework	Yes, core	Python	Apache 2.0	~7k	current
10	UUnstructured	Library	No	Python	Apache 2.0	~12k	current
11	R2R SciPhi	Platform	Yes	Python	MIT	~7k	current
12	txtai NeuML	Framework	Yes	Python	Apache 2.0	~11k	current
13	LLMWare	Framework	No	Python	Apache 2.0	~13k	current
14	Verba Weaviate	Application	No	Python	BSD-3	~7k	current

#1

LlamaIndex

LlamaIndex

Ingestion

and retrieval

• Deep document handling and many indexing strategies
• The common advice is LlamaIndex for retrieval, LangGraph for orchestration

Best forPDFs, knowledge bases, structured data

#2

LangChain & LangGraph

LangChain

Orchestration

first

• LangGraph adds durable state and checkpointing
• Widely used, and widely argued about

Best foragents that retrieve as one step among many