GraphRAG

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

来源语言 · EN

先看实际效果

产品界面预览

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
查看完整图片
许可

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
产品界面预览

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
查看完整图片
许可

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
产品界面预览

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
查看完整图片
许可

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

产品介绍

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.

来源语言 · EN
主要功能

GraphRAG 能做什么

  1. 01Index text into graph-based representations
  2. 02Query indexed collections
  3. 03Tune prompts for a dataset
来源语言 · EN
差异特点

它有什么不同

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.

来源语言 · EN
适合人群

谁值得考虑

  • Researchers exploring graph-based retrieval
  • Developers comparing retrieval approaches
来源语言 · EN
选择之前

需要核对的事项

  • 用一个真实任务试用,并对照自己的要求检查结果。
  • 核实服务会保存、共享或用于训练哪些数据。
  • 在官网确认当前价格、使用限额和取消条款。
官网核验日期

官网链接核查于 2026年11月9日;功能声明尚未经过独立测试。