Skip to main content

Agentic GraphRAG

agrag is a graph-based retrieval-augmented generation library with agentic reasoning. It builds a knowledge graph from corpus text and answers questions with multi-strategy retrieval and LLM reasoning.

Installation​

uv pip install agentic-graphrag
import agrag

print(agrag.__version__)

Where to go next​

  • Ingest documents — add files, directories, raw text, or prebuilt documents to a graph, and handle per-source errors.
  • Configure storage backends — set up an embedder, a vector store, and a graph store.
  • Extract and resolve entities — pull entities and relations out of chunks, and group duplicate mentions.
  • API Reference — generated from the agrag package's docstrings. Rebuilt on every docs build; run make docs-api to regenerate it locally.

Guides land here as each part of the library ships.