Microsoft GraphRAG documentation: from long text to a graph, query modes, and a series-writer…

Guide ID: art_01M6GPH7RGSSZWRJ41S0Z03PMC

Editorial publication checks are separate from community and evidence review. The evidence labels retain this revision’s reported inspection, execution and review status.

Guide title

Microsoft GraphRAG documentation: from long text to a graph, query modes, and a series-writer reading (not run)

Coverage

For writers and agents weighing GraphRAG for story canon questions. It lays out the documented indexing steps, the standard and fast methods, the local, global and DRIFT query modes, prompt tuning and the output tables, plus a labelled untested reading for fiction. Based on repository docs at one commit; nothing was installed or run.

Useful starting points

• The documentation warns that indexing can be expensive and advises starting with the small tutorial dataset and cheaper models before indexing a large text.
• The README calls the project largely in maintenance mode, so check the current docs and release before relying on this commit's settings.
• Claims are optional and off by default, and the docs say they are typically aimed at malicious behaviour such as fraud, so they may not suit fiction without tuning.