GraphRAG

GraphRAG is a research toolkit for extracting graph-based structures from text to support retrieval and question answering

출처 원문 · EN

실제 예시 보기

제품 화면

Experimental GraphRAG demo: dataset and search controls

Author screenshot of the experimental, unsupported Unified Search demo with A Christmas Carol selected and local/global controls. This is its initial configuration view, not generated answer evidence.

Experimental GraphRAG demo: dataset and search controls
전체 이미지 보기
라이선스

Author PNG copied at original dimensions and re-encoded losslessly with exact decoded pixels; no crop, resize, added UI or invented answers. The pinned unified-search-app README identifies these states and says the demo is experimental/unsupported. The visible browser chrome and partial viewport are retained.

    MIT License

    Copyright (c) Microsoft Corporation.

    Permission is hereby granted, free of charge, to any person obtaining a copy
    of this software and associated documentation files (the "Software"), to deal
    in the Software without restriction, including without limitation the rights
    to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
    copies of the Software, and to permit persons to whom the Software is
    furnished to do so, subject to the following conditions:

    The above copyright notice and this permission notice shall be included in all
    copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
    SOFTWARE
제품 화면

Experimental GraphRAG demo: select a suggested question

The author’s demo shows suggested questions and a selected Marley question while the local/global searches are still generating. No completed comparison answer or citation quality is asserted.

Experimental GraphRAG demo: select a suggested question
전체 이미지 보기
라이선스

Author PNG copied at original dimensions and re-encoded losslessly with exact decoded pixels; no crop, resize, added UI or invented answers. The pinned unified-search-app README identifies these states and says the demo is experimental/unsupported. The visible browser chrome and partial viewport are retained.

    MIT License

    Copyright (c) Microsoft Corporation.

    Permission is hereby granted, free of charge, to any person obtaining a copy
    of this software and associated documentation files (the "Software"), to deal
    in the Software without restriction, including without limitation the rights
    to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
    copies of the Software, and to permit persons to whom the Software is
    furnished to do so, subject to the following conditions:

    The above copyright notice and this permission notice shall be included in all
    copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
    SOFTWARE
제품 화면

Experimental GraphRAG demo: inspect a community report

The author’s demo shows community selection, a partial entity-graph viewport and the visible portion of a selected Scrooge report. It is an interface preview of the experimental app, not a complete answer or site benchmark.

Experimental GraphRAG demo: inspect a community report
전체 이미지 보기
라이선스

Author PNG copied at original dimensions and re-encoded losslessly with exact decoded pixels; no crop, resize, added UI or invented answers. The pinned unified-search-app README identifies these states and says the demo is experimental/unsupported. The visible browser chrome and partial viewport are retained.

    MIT License

    Copyright (c) Microsoft Corporation.

    Permission is hereby granted, free of charge, to any person obtaining a copy
    of this software and associated documentation files (the "Software"), to deal
    in the Software without restriction, including without limitation the rights
    to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
    copies of the Software, and to permit persons to whom the Software is
    furnished to do so, subject to the following conditions:

    The above copyright notice and this permission notice shall be included in all
    copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
    SOFTWARE

제품 소개

GraphRAG is a structured, hierarchical approach to Retrieval Augmented Generation (RAG), as opposed to naive semantic-search approaches using plain text snippets. The GraphRAG process involves extracting a knowledge graph out of raw text, building a community hierarchy, generating summaries for these communities, and then leveraging these structures when perform RAG-based tasks.

출처 원문 · EN
출처에 기재된 기능

GraphRAG의 주요 기능

  1. 01Index text into graph-based representations
  2. 02Query indexed collections
  3. 03Tune prompts for a dataset
출처 원문 · EN
차별점

차별점

The repository is in maintenance mode and is a research demonstration rather than an officially supported Microsoft service. Indexing can be costly and should start with a small dataset.

출처 원문 · EN
추천 대상

추천 대상

  • Researchers exploring graph-based retrieval
  • Developers comparing retrieval approaches
출처 원문 · EN
선택하기 전에

확인할 사항

  • 실제 작업 하나를 시험하고 결과가 요구 사항에 맞는지 비교합니다.
  • 어떤 데이터가 저장·공유되거나 학습에 사용되는지 확인합니다.
  • 공식 사이트에서 현재 가격, 사용 한도, 해지 조건을 확인합니다.
마지막 확인

공식 링크는 2026년 11월 9일에 확인했습니다. 기능 주장은 독립적으로 테스트하지 않았습니다.