GraphRAG

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

出典の原文 · EN

実際の例を見る

製品画面

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
画像全体を見る
ライセンス

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
製品画面

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
画像全体を見る
ライセンス

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
製品画面

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
画像全体を見る
ライセンス

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

製品紹介

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.

出典の原文 · EN
出典に記載された機能

GraphRAGでできること

  1. 01Index text into graph-based representations
  2. 02Query indexed collections
  3. 03Tune prompts for a dataset
出典の原文 · EN
違い

他と異なる点

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.

出典の原文 · EN
おすすめ

検討に向く人

  • Researchers exploring graph-based retrieval
  • Developers comparing retrieval approaches
出典の原文 · EN
選ぶ前に

確認すること

  • 実際の作業を一つ試し、要件に合うか比較する。
  • 保存・共有・学習に利用されるデータを確認する。
  • 公式サイトで現在の料金、利用制限、解約条件を確認する。
サイト確認日

公式リンクは2026年11月9日に確認しました。機能の主張は独立して検証していません。