GraphRAG

什么是 GraphRAG？

GraphRAG is a research project and data pipeline that uses language models to extract structured information from unstructured text. It explores knowledge-graph memory structures as a way to form targeted context for question answering, particularly when working with private data.

The repository presents a methodology and demonstration code, not an officially supported Microsoft offering. The project is largely in maintenance mode, with no new features or pull requests planned; bug fixes and dependency updates may be made as appropriate.

GraphRAG 能做什么？

LLM-based text transformation

The pipeline is designed to extract meaningful, structured data from unstructured text using language models.

Graph-based context for questions

Knowledge-graph memory structures are used to form targeted context intended to help language models answer questions about a data collection.

Command-line quickstart

The README directs users to a command-line quickstart as the entry point for trying the system.

Prompt tuning guidance

The project recommends fine-tuning prompts for the user's data because default prompts may not produce the best results.

Documented research methodology

The repository and linked documentation explain an approach to using graph structures to enhance LLM outputs; the code is presented as a demonstration.

使用场景

“Explore question answering over private data”

Evaluate whether graph-structured context can help a language model reason about an organization's own text collection; the project explicitly frames private-data question answering as a target.

“Prototype an LLM data pipeline”

Use the demonstration pipeline to investigate transforming unstructured text into structured data with language models.

“Study graph-based RAG methods”

Researchers and developers can examine the repository's methodology for applying knowledge-graph memory structures to LLM outputs.

常见问题

What does GraphRAG do?

It is a data pipeline and transformation suite designed to extract structured information from unstructured text with language models, and a research method for using graph structures to provide context for question answering.

How should I get started?

The README recommends following the command-line quickstart and reading the documentation. It also recommends starting small because indexing can be expensive.

Is GraphRAG an officially supported Microsoft product?

No. The repository says the code is a demonstration and is not an officially supported Microsoft offering.

Is the project actively adding features?

The README says GraphRAG is largely in maintenance mode and will not accept new pull requests or implement new features. Bug fixes and dependency updates may be made as appropriate.

Should I use the default prompts?

The project recommends prompt tuning for your data, noting that using GraphRAG out of the box may not yield the best results.

快速信息