Struct GraphRAG 

pub struct GraphRAG<M: BaseChatModel> { /* private fields */ }

GraphRAG: Knowledge Graph-based Retrieval Augmented Generation.

Wraps an LLM for entity/relation extraction and community summarization, and an in-memory GraphStore for graph operations.

Implementations

impl<M: BaseChatModel> GraphRAG<M>

pub fn new(llm: M) -> Self

Creates a new GraphRAG instance with the given LLM.

pub fn with_config(self, config: GraphRAGConfig) -> Self

Sets a custom configuration.

pub async fn add_documents( &self, docs: &[Document], ) -> Result<(), GraphRAGError>

Adds documents to the knowledge graph by extracting entities and relations from each document via the LLM.

pub async fn build_communities(&self) -> Result<(), GraphRAGError>

Runs hierarchical Leiden community detection and generates one summary per community via the LLM.

Level-0 communities are summarized from their entities and internal relations; each higher level is rolled up from its child-community summaries plus the relations crossing the children. Community ids index the summary vector one-to-one.

pub async fn query( &self, q: &str, mode: QueryMode, ) -> Result<GraphRAGResult, GraphRAGError>

Queries the knowledge graph using the specified mode.

pub async fn entity_count(&self) -> usize

Returns the number of entities in the graph.

pub async fn relation_count(&self) -> usize

Returns the number of relations in the graph.

pub async fn community_count(&self) -> usize

Returns the number of communities across all hierarchy levels.

pub async fn communities(&self) -> Vec<Community>

Returns a clone of the detected community hierarchy (empty until GraphRAG::build_communities has run).

pub async fn community_summaries(&self) -> Vec<String>

Returns one summary per community, indexed by Community::id.