VertexRAG

VertexRAG — Managed Graph RAG API for Knowledge-Aware Retrieval

VertexRAG is a managed Graph-RAG API that turns your documents into a knowledge graph of entities and relationships, then answers queries with hybrid vector + graph retrieval. Every answer is pinned to its source chunk and document ID. The platform is self-hosted on physically isolated servers in Hyderabad, India, with no third-party LLM sharing and no training on your data.

Graph RAG outperforms pure vector search on multi-hop questions, entity disambiguation, and structured filtering — the cases where retrieval typically fails. Read the API documentation to ingest your first corpus in five minutes, or try the interactive playground.

How VertexRAG works

Three stages transform flat documents into a queryable knowledge graph:

1. Ingest

Push any document via REST or SDK — PDFs, Markdown, HTML. Chunked adaptively, entities extracted, embeddings and graph triples stored in one call. Learn more in our explainer: What is Graph RAG? A 5-minute explainer.

2. Build knowledge graph

Entities and relationships become typed graph nodes. Multi-hop paths enable reasoning that flat vector search cannot reach. Read the deep dive: Multi-hop reasoning: the real edge of graph-powered RAG.

3. Retrieve

Queries fuse dense vector similarity with graph-path scores, then rerank — every answer pinned to its source chunk and doc ID. Compare with vector-only retrieval: Graph RAG vs Vector RAG: when each actually wins.

Free RAG developer tools

  • RAG Evaluation — score retrieval quality with precision, recall, and faithfulness signals.
  • LLM Token Counter — estimate tokens and cost instantly for OpenAI and Google Gemini.
  • AI Entity Extractor — pull entities and visualise them as an interactive knowledge graph.
  • Text Chunker for RAG — preview chunking strategies for vector and graph retrieval pipelines.
  • Readability Analyzer — measure text complexity, reading time, and AI optimisation potential.

Frequently asked questions

What is graph RAG?

Graph RAG is retrieval-augmented generation that uses a knowledge graph of extracted entities and relationships alongside vector embeddings, so queries can traverse connections that vector similarity alone misses.

How is VertexRAG different from Microsoft GraphRAG?

Microsoft’s GraphRAG is an open-source Python library focused on community detection and hierarchical summarisation. VertexRAG is a hosted REST API with first-class multi-hop traversal, incremental ingest, and predictable per-query pricing. Read the full comparison: Graph RAG vs Microsoft GraphRAG.

Is my data shared with LLM providers?

No. VertexRAG is entirely self-hosted. Data is stored on physically isolated servers in Hyderabad, India and never sent to third-party LLM providers for training. See our self-hosted architecture write-up for details.

How do I evaluate a RAG pipeline?

Measure precision, recall, and faithfulness per query. Our RAG evaluation guide walks through the four metrics and a minimal Python loop. Score your own content instantly with the free RAG Evaluation tool.

Pricing

VertexRAG uses simple credit-based pricing — no subscriptions, credits never expire. Starter is $10 for 2,000 credits, Pro is $30 for 5,000 credits, Scale is $50 for 8,000 credits. See full pricing for the complete breakdown.

References and further reading