Enhance vector search with graph structures. Composable pipeline, pluggable storage.
import { createGraph, query } from "graphrag-sdk";
import { lexicalGraph, chunkEntityGeneration, naiveSearch } from "@graphrag-sdk/lexical";
import { inMemoryGraph, inMemoryVector, inMemoryKV } from "@graphrag-sdk/in-memory-storage";
import { openai } from "@ai-sdk/openai";
const graph = createGraph({
graph: lexicalGraph({
graphStoreFactory: inMemoryGraph,
vectorStoreFactory: inMemoryVector,
kvStoreFactory: inMemoryKV,
pipeline: [chunkEntityGeneration()],
}),
model: openai("gpt-4o-mini"),
embedding: openai.embedding("text-embedding-3-small"),
});
await graph.insert("Your documents...");
const { text } = await query({ graph, search: naiveSearch(), prompt: "Your question?" });Build your graph structure with composable processors:
pipeline: [
chunkEntityGeneration(), // chunks → entities + relationships
entityCommunityGeneration(), // entities → Leiden communities + reports
similarChunkConnection(), // chunks ↔ similarity edges
hypotheticalQuestionGeneration(), // chunks → question embeddings
]Each strategy retrieves context differently:
| Strategy | What it searches | Best for |
|---|---|---|
naiveSearch() | Chunk embeddings | Simple fact lookups |
entityEnhancedSearch() | Chunks + entity graph traversal | Relational context across chunks |
localCommunitySearch() | Entity neighborhoods + communities | Entity-specific questions |
globalCommunitySearch() | Community reports | Thematic/summary questions |
similarChunkTraversalSearch() | Chunk similarity graph | Connected context |
hypotheticalQuestionSearch() | Question embeddings | Q&A retrieval |
Pre-configured pipelines matching well-known implementations:
import { microsoftGraphragPreset, lightragPreset } from "@graphrag-sdk/lexical";
// Microsoft GraphRAG — entities + communities
lexicalGraph({ ...microsoftGraphragPreset(), graphStoreFactory, vectorStoreFactory, kvStoreFactory })
// LightRAG — entities only
lexicalGraph({ ...lightragPreset(), graphStoreFactory, vectorStoreFactory, kvStoreFactory })Pure vector search retrieves similar chunks. Graph structures add: