Knowledge Graph

Overview

The knowledge graph is the foundation of test data generation. It's a structured representation of your domain knowledge with nodes (documents and chunks) and relationships (connections between concepts).

Unlike raw text, a knowledge graph enables:

• Semantic search through embeddings
• Multi-hop reasoning across related chunks
• Understanding of document structure and relationships
• Efficient retrieval for test generation

Creating a Knowledge Graph

Start with document nodes representing your source content:

import { graph, DocumentNode } from '@open-evals/generator'

const documents = [
  new DocumentNode(
    'typescript-basics.md', // Unique identifier
    'TypeScript is a typed...', // Content
    { category: 'tutorial' } // Custom metadata
  ),
  new DocumentNode('advanced-types.md', 'Advanced TypeScript features...', {
    category: 'reference',
  }),
]

const knowledgeGraph = graph(documents)

At this point, your graph contains only document nodes. You'll enrich it with transforms.

Transform Pipeline

Transforms are composable functions that enrich your graph. Chain them together to build sophisticated knowledge representations:

import {
  transform,
  chunk,
  embed,
  relationship,
  summarize,
} from '@open-evals/generator'
import { RecursiveCharacterSplitter } from '@open-evals/rag'
import { openai } from '@ai-sdk/openai'

const enrichedGraph = await transform(graph(documents))
  .pipe(summarize(openai.chat('gpt-4o')))
  .pipe(chunk(new RecursiveCharacterSplitter({ chunkSize: 512 })))
  .pipe(embed(openai.embedding('text-embedding-3-small')))
  .pipe(relationship())
  .apply()

Each transform runs sequentially, building on the previous step's output.

Next Steps

• Transforms - Learn about all available transforms and how to use them
• Usage - Query, traverse, and modify your knowledge graph
• Advanced - Persistence, custom transforms, and best practices