Crate graphrag_core 

GraphRAG Core

Portable core library for GraphRAG - works on both native and WASM platforms.

This is the foundational crate that provides:

• Knowledge graph construction and management
• Entity extraction and linking
• Vector embeddings and similarity search
• Graph algorithms (PageRank, community detection)
• Retrieval systems (semantic, keyword, hybrid)
• Caching and optimization

Platform Support

• Native: Full feature set with optional CUDA/Metal GPU acceleration
• WASM: Browser-compatible with Voy vector search and Candle embeddings

Feature Flags

• wasm: Enable WASM compatibility (uses Voy instead of HNSW)
• cuda: Enable NVIDIA GPU acceleration via Candle
• metal: Enable Apple Silicon GPU acceleration
• webgpu: Enable WebGPU acceleration for browser (via Burn)
• pagerank: Enable PageRank-based retrieval
• lightrag: Enable LightRAG optimizations (6000x token reduction)
• caching: Enable intelligent LLM response caching

Quick Start

use graphrag_core::{GraphRAG, Config};

let config = Config::default();
let mut graphrag = GraphRAG::new(config)?;
graphrag.initialize()?;

Re-exports

pub use crate::config::Config;

pub use crate::core::ChunkId;

pub use crate::core::Document;

pub use crate::core::DocumentId;

pub use crate::core::Entity;

pub use crate::core::EntityId;

pub use crate::core::EntityMention;

pub use crate::core::ErrorContext;

pub use crate::core::ErrorSeverity;

pub use crate::core::ErrorSuggestion;

pub use crate::core::GraphRAGError;

pub use crate::core::KnowledgeGraph;

pub use crate::core::Relationship;

pub use crate::core::Result;

pub use crate::core::TextChunk;

pub use crate::core::traits::Embedder;

pub use crate::core::traits::EntityExtractor;

pub use crate::core::traits::GraphStore;

pub use crate::core::traits::LanguageModel;

pub use crate::core::traits::Retriever;

pub use crate::core::traits::Storage;

pub use crate::core::traits::VectorStore;

pub use crate::storage::MemoryStorage;

pub use crate::builder::GraphRAGBuilder;

Modules

async_graphrag

Async GraphRAG implementation Async GraphRAG System

async_processing

Async processing pipelines Async processing utilities for GraphRAG operations

builder

Builder pattern implementations GraphRAG builder module

config

Configuration management and loading Configuration types for GraphRAG.

core

Core traits and types Core data structures and abstractions for GraphRAG

critic

RAG answer evaluation and criticism LLM-based critique of generated answers.

embeddings

Embedding generation and providers Embedding generation for GraphRAG

entity

Entity extraction and management Entity and relationship extraction.

evaluation

Evaluation framework for query results and pipeline validation Evaluation framework for GraphRAG system

generation

Text generation and LLM interactions (async feature only) Answer generation from retrieval results.

graph

Graph data structures and algorithms Knowledge-graph construction and graph algorithms.

inference

Inference module for model predictions Implicit relationship inference system

monitoring

Monitoring, benchmarking, and performance tracking Monitoring and benchmarking utilities for GraphRAG

nlp

Natural language processing utilities Advanced NLP Module

ollama

Ollama LLM integration Ollama LLM integration

optimization

Graph optimization (weight optimization, DW-GRPO) Graph optimization module

parallel

Parallel processing utilities for GraphRAG

persistence

Persistence layer for knowledge graphs (workspace management always available) Persistence layer for GraphRAG knowledge graphs

pipeline

Data processing pipelines Pipeline Module

pipeline_executor

Composable pipeline executor for build-graph operations Composable pipeline executor for GraphRAG operations

prelude

Prelude module containing the most commonly used types

query

Query processing and execution Query understanding for hierarchical GraphRAG.

reranking

Reranking utilities for improving search result quality Reranking module for improving retrieval accuracy

retrieval

Retrieval strategies and implementations Retrieval system: finds the chunks and entities relevant to a query.

storage

Storage backends and persistence Storage layer for GraphRAG

summarization

Text summarization capabilities Summarization of chunks and communities, and aggregation of query results.

text

Text processing and chunking Text processing pipeline: chunking, enrichment, and analysis.

vector

Vector operations and embeddings Vector store and similarity search.

Macros

config_error

Helper macros for creating specific error types

generation_error

Creates a generation error with a message

retrieval_error

Creates a retrieval error with a message

storage_error

Creates a storage error with a message

Structs

GraphRAG

This is the primary entry point for using GraphRAG. It orchestrates all components: knowledge graph, retrieval, generation, and caching.