GraphRAG

3.0.9 verified Fri May 01 auth: no python

GraphRAG is a graph-based retrieval-augmented generation (RAG) system by Microsoft. The current version is 3.0.9, with frequent patch releases. Requires Python >=3.11, <3.14.

pip install graphrag

Common errors

error ModuleNotFoundError: No module named 'graphrag' ↓

cause The library is not installed or Python environment is wrong.

fix

Run pip install graphrag and ensure you are using Python >=3.11.

error AttributeError: module 'graphrag' has no attribute 'Indexer' ↓

cause Import path changed in v3.0.0; Indexer is now in graphrag.index.

fix

Use from graphrag.index import Indexer instead of from graphrag import Indexer.

error OSError: [Errno 2] No such file or directory: 'config.yaml' ↓

cause Missing configuration file; GraphRAG requires a config file to run indexing or queries.

fix

Run graphrag init --force in your project directory to create default config.yaml.

error ValueError: Unknown embedding model: ... ↓

cause The embedding model specified in config is not recognized or not installed.

fix

Check the 'embedding.model' field in config.yaml and ensure you have the correct model name (e.g., text-embedding-ada-002).

Warnings

breaking v3.0.0 introduced a monorepo restructure: many import paths changed. Use `from graphrag.index` instead of `from graphrag`. Old imports will break. ↓

fix Update imports to match new package structure (e.g., graphrag.index, graphrag.query, graphrag.cache, etc.)

deprecated The `QueryEngine` from `graphrag.query_engine` has been moved to `graphrag.query`. The old path may be removed in a future release. ↓

fix Use `from graphrag.query import QueryEngine`

gotcha The `litellm` dependency was compromised in the past; GraphRAG pins a specific safe version. Do not upgrade litellm independently. ↓

fix Keep litellm pinned via GraphRAG's requirements; do not override.

breaking Python 3.13 support was added in v3.0.2, but v3.0.0-v3.0.1 did not support it. Python 3.11 is the minimum. ↓

fix Upgrade to v3.0.2+ for Python 3.13 compatibility.

gotcha The `graphrag init --force` command is required to reinitialize config after upgrading to v3.0.0; otherwise configs may be incompatible. ↓

fix Run `graphrag init --force` after upgrading to regenerate config.

Imports

Indexer
wrong
```
from graphrag import Indexer
```
correct
```
from graphrag.index import Indexer
```
In v3.0.0+, the index package was restructured; direct import from graphrag is deprecated.

QueryEngine

wrong

from graphrag.query_engine import QueryEngine

correct

from graphrag.query import QueryEngine

Renamed in v3.0.0 from query_engine to query.

run_pipeline
wrong
```
from graphrag.pipeline import run
```
correct
```
from graphrag.index.run import run_pipeline
```
Pipeline module restructured; use run_pipeline from index.run.

Quickstart

Minimal import and setup for GraphRAG. Note: Full indexing requires a config file and data directory; run `graphrag init --force` first.

import os
os.environ.pop("OPENAI_API_KEY", None)  # placeholder; set your key via env var
from graphrag.index import Indexer
from graphrag.query import QueryEngine
# Index documents (requires a data directory and config)
# Run: graphrag init --force to create default config
print("GraphRAG ready")