Skip to content
CAI
Software that uses CAICheck a score

microsoft/graphrag

62.7

Weak · 3 November 2026

32.7k

lines of production code

Python

primary language

4

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

Features

Added example notebooks for input and storage packages

New example notebooks have been added to the \graphrag-input\ and \graphrag-storage\ packages to demonstrate usage. The \graphrag-input\ example shows how to configure input readers for CSV and MarkItDown (PDF) sources, including handling nested JSON properties. The \graphrag-storage\ examples demonstrate basic file storage operations and how to register and use custom storage implementations via the factory pattern.

_packages/graphrag-input/example\_notebooks, packages/graphrag-storage/example\notebooks · high confidence

Introduce GraphRAG unified search application

The unified-search-app/app directory now contains a complete Streamlit-based interface for exploring GraphRAG-indexed data. Users can load datasets from local storage or Azure Blob Storage, then run and compare four distinct search modes—Basic RAG, Local Search, Global Search, and Drift Search—side by side. The app also supports generating suggested questions, browsing community reports, and viewing detailed citations with hyperlinked entities and relationships.

unified-search-app/app · high confidence

Introduce graphrag-llm package with completion, embedding, and configuration modules

The new graphrag-llm package provides a unified interface for language model interactions, featuring dedicated modules for text completion and embedding generation. Users can configure models via Pydantic-based settings (ModelConfig) that support Azure OpenAI and other providers, with built-in handling for authentication methods like API keys and Managed Identity. The package includes factory functions to create completion and embedding clients, supporting both streaming and non-streaming responses, along with utilities for structured response validation. Example notebooks and documentation are provided to demonstrate basic usage patterns for completions and embeddings.

packages/graphrag-llm · high confidence

Introduce unified TableProvider abstraction with native Cosmos DB, CSV, and Parquet backends

The storage layer now uses a new \TableProvider\ interface to manage table storage, enabling row-by-row streaming access via the \Table\ abstraction. This change introduces three concrete implementations: a native \CosmosTableProvider\ that stores data directly in Azure Cosmos DB (NoSQL API) using namespace-based partitioning and eliminating the previous Parquet serialization round-trip; a \CSVTableProvider\ for file-based storage using CSV format; and a \ParquetTableProvider\ for Parquet-based storage. A \TableProviderFactory\ allows selecting the backend via configuration (e.g., \cosmosdb\, \csv\, \parquet\), and the Cosmos implementation includes a legacy container fallback to support migration from older storage formats.

_packages/graphrag-storage/graphrag\storage/tables · high confidence

Introduces core data models and vector store infrastructure for GraphRAG indexing

This change establishes the foundational data structures and storage interfaces for the GraphRAG indexing pipeline. It adds Pydantic-based data models for key entities including Documents, TextUnits, Entities, Relationships, Communities, and Covariates, providing a standardized schema for indexing results. Additionally, it introduces a new \graphrag-vectors\ package that defines a generic vector store abstraction with support for filtering expressions, timestamp handling, and schema validation, alongside a factory pattern in \graphrag-common\ to manage service instantiation. These components provide the necessary type definitions and storage contracts for subsequent indexing operations like embedding and claim extraction.

graphrag-monorepo · high confidence

New graphrag-chunking and graphrag-common packages with example notebooks

This change introduces two new packages: graphrag-chunking, which provides text chunking capabilities (SentenceChunker and TokenChunker) via a configuration-driven factory, and graphrag-common, which offers a flexible dependency injection factory and a configuration loading system with Pydantic support. Both packages now include example Jupyter notebooks demonstrating their usage, and the nltk bootstrap process has been updated to download additional required resources (punkt\_tab, averaged\_perceptron\_tagger\_eng, maxent\_ne\_chunker\_tab).

packages/graphrag-chunking, packages/graphrag-common · high confidence

New graphrag-storage package with unified storage interface

A new \graphrag-storage\ package has been introduced, providing a unified \Storage\ interface and factory for managing data persistence. This change consolidates storage implementations, including \FileStorage\, \MemoryStorage\, \AzureBlobStorage\, and \AzureCosmosStorage\, under a common API. Users can now configure storage backends via \StorageConfig\ and create instances through \create\_storage\, supporting both local file systems and Azure cloud services with namespace isolation capabilities.

_packages/graphrag-storage/graphrag\storage · high confidence

New graphrag-vectors package with multi-backend vector store support

The new \graphrag-vectors\ package provides vector store implementations for GraphRAG, supporting LanceDB, Azure AI Search, and Azure Cosmos DB backends. It introduces a factory-based creation API (\create\_vector\_store\, \vector\_store\_factory\) and a configuration-driven approach (\VectorStoreConfig\, \IndexSchema\) for managing vector indices. The package includes a generic filtering system using the \F\ builder for metadata filtering (compiled to OData for Azure AI Search and SQL for Cosmos DB), timestamp handling, and CRUD operations for documents. Example notebooks are provided to demonstrate basic usage, factory patterns, and custom vector store registration.

packages/graphrag-vectors · high confidence

New modular input reader package for GraphRAG

The \graphrag\_input\ package has been introduced to standardize how documents are loaded into the pipeline. It provides a factory-based system (\InputReaderFactory\) that supports multiple file formats via specific readers: CSV, JSON, JSON Lines, Parquet, plain text, and Markdown (via MarkItDown). The package includes configuration models (\InputConfig\) to define file patterns, encodings, and column mappings, as well as base classes (\InputReader\, \StructuredFileReader\) that handle async iteration and document generation. This change establishes the foundational infrastructure for input handling, including utilities for hashing and property retrieval, replacing or abstracting previous ad-hoc loading logic.

_packages/graphrag-input/graphrag\input · high confidence

New public Python API and CLI for indexing and querying

GraphRAG now exposes a structured Python API and a command-line interface (CLI) for programmatic use. The new \graphrag.api\ module provides \build\_index\ for running indexing pipelines (including streaming and update runs) and query functions such as \global\_search\, \local\_search\, and \drift\_search\ (with streaming variants), along with \generate\_indexing\_prompts\ for prompt tuning. The CLI, built with Typer, adds \init\, \index\, \update\, \query\, and \prompt-tune\ subcommands, enabling users to initialize projects, run indexing with configurable methods and caching, execute searches with streaming output, and generate tuning prompts from the command line.

packages/graphrag/graphrag · high confidence

New standalone GraphRAG cache package with factory pattern and SQLite support

The \graphrag-cache\ package introduces a new, standalone caching system for GraphRAG, replacing the previous \pipeline\_cache\ implementation. It provides a factory-based approach (\create\_cache\) that supports four built-in cache types: \Json\, \Memory\, \Noop\, and \SQLite\. The new \SQLiteCache\ offers concurrency-safe, namespaced storage using WAL mode, while \JsonCache\ and \MemoryCache\ have been refactored to use the new \Storage\ abstraction. Users can now register custom cache implementations via \register\_cache\ and configure storage backends through \CacheConfig\. Example notebooks demonstrate both basic usage and custom cache extension.

packages/graphrag-cache · high confidence

Removals

Entity extraction and description summarization helper verbs removed

The \entity\_extract\ and \summarize\_descriptions\ indexing verbs, along with their underlying implementation strategies (including \graph\_intelligence\ and \nltk\ for extraction, and \graph\_intelligence\ for summarization), have been deleted from the \graphrag/index/verbs/entities\ package. This change removes the ability to extract entities from text and summarize entity/relationship descriptions using these specific helper verbs.

graphrag/index/verbs/entities · high confidence

Removal of OpenAI LLM implementation

The OpenAI LLM implementation module (graphrag.llm.openai) has been removed. This includes the deletion of all associated files such as client creation, factory functions, configuration classes, and specific LLM wrappers (chat, completion, embeddings, JSON parsing, history tracking, and token replacing). This change replaces the previous OpenAI-specific logic with a different implementation (fnllm) as indicated by the commit message.

graphrag/llm/openai · high confidence

Removal of custom indexing verb overrides

The custom override implementations for the aggregate, concat, and merge indexing verbs have been removed from the codebase. This deletion eliminates the specialized logic that previously handled specific aggregation operations, column-wise versus row-wise concatenation, and JSON-based merging strategies, effectively reverting these operations to their standard or upstream definitions.

graphrag/index/verbs/overrides · high confidence

Removal of graph layout verb and strategies

The \layout\_graph\ verb and its associated UMAP and zero-position strategies have been removed from the indexing pipeline. This change eliminates the capability to apply graph layout algorithms to position nodes based on embeddings or structural properties, effectively removing this step from the graph processing workflow.

graphrag/index/verbs/graph/layout · high confidence

Removal of graphrag graph merge helper verbs

The graph merge helper verbs located in the \graphrag/index/verbs/graph/merge\ directory have been removed. This includes the \merge\_graphs\ verb, which previously combined multiple graphs into a single graphml output, along with its supporting default configurations (node/edge operations, concatenation separators) and type definitions. Users relying on this specific graph merging functionality will no longer have access to these components.

graphrag/index/verbs/graph/merge · high confidence

Removal of graphrag.query.input module

The graphrag.query.input module has been removed from the codebase. This deletion eliminates the package-level initialization file that previously defined the GraphRAG Orchestration Inputs namespace, meaning any code relying on imports from this specific path will no longer function.

_examples\notebooks, graphrag/query/input · high confidence

Removal of indexing engine example code

The \examples\ directory has been completely removed, deleting all demonstration code for the indexing engine. This includes the main \README.md\ and all subdirectories (\custom\_input\, \custom\_set\_of\_available\_verbs\, \custom\_set\_of\_available\_workflows\, \entity\_extraction\, \interdependent\_workflows\, \multiple\_workflows\, \single\_verb\, \use\_built\_in\_workflows\, and \various\_levels\_of\_configs\). Users will no longer have access to these reference implementations for running pipelines via Python API or configuration files.

examples · high confidence

Removal of internal LLM base implementations

The \graphrag/llm/base\ module has been removed, deleting the \BaseLLM\, \CachingLLM\, and \RateLimitingLLM\ classes along with their cache-key generation utilities. This change eliminates the internal base implementations and caching/rate-limiting wrappers that were previously exposed from this package location, indicating a shift in how LLM interactions are structured or delegated elsewhere in the system.

graphrag/llm/base · high confidence

Removal of internal LLM limiting and error modules

The \graphrag.llm\ package and its \limiting\ subpackage have been removed, deleting all internal LLM limiter implementations (including \LLMLimiter\, \TpmRpmLLMLimiter\, \CompositeLLMLimiter\, and \NoopLLMLimiter\), the \create\_tpm\_rpm\_limiters\ factory, and the \RetriesExhaustedError\ class. This change eliminates the built-in token-per-minute and requests-per-minute rate-limiting logic and associated error handling from this location, as the package is being replaced by \fnllm\.

graphrag/llm, graphrag/llm/limiting · high confidence

Removal of legacy CSV and text input loaders

The legacy input loading implementations for CSV and text files have been removed from the \graphrag/index/input\ module. This change deletes the specific loaders (\csv.py\, \text.py\) and the central dispatch logic (\load\_input.py\) that previously handled these file types, indicating a shift away from supporting these specific input formats in this code path.

graphrag/index/input · high confidence

Removal of legacy DataFrame input loaders

The legacy input loader module located at graphrag/query/input/loaders has been removed. This deletion eliminates the previous mechanism for converting pandas DataFrames into GraphRAG model objects (such as Entities, Relationships, and Covariates) and discards the associated utility functions for column type validation and conversion. Users relying on these specific DataFrame-based loading paths will need to adopt the updated input handling methods provided by the current version.

graphrag/query/input/loaders · high confidence

Removal of legacy GlobalSearch and LocalSearch implementations

The legacy structured search components in \graphrag/query/structured\_search/global\_search\ and \graphrag/query/structured\_search/local\_search\ have been removed. This includes the deletion of the \GlobalSearch\ and \LocalSearch\ orchestration classes, their associated system prompts (\map\_system\_prompt.py\, \reduce\_system\_prompt.py\, \system\_prompt.py\), the \GlobalCommunityContext\ builder, and the \GlobalSearchLLMCallback\ handler. Users relying on these specific classes for query execution will need to migrate to the updated search architecture.

_graphrag/query/structured\_search/global\search · high confidence

Removal of legacy GraphRAG knowledge model package

The entire \graphrag.model\ package has been removed, deleting the legacy data model definitions and protocols. This includes the deletion of the package root (\\_\init\\_.py\) and all constituent model files: \community.py\, \community\_report.py\, \covariate.py\, \document.py\, \entity.py\, \identified.py\, \named.py\, \relationship.py\, \text\_unit.py\, and \types.py\. Users relying on these specific data classes for direct integration or analytics will need to migrate to the new model structure.

graphrag/model · high confidence

Removal of legacy LLM loading utilities

The \graphrag/index/llm\ module has been removed, deleting the \load\_llm\ and \load\_llm\_embeddings\ functions along with their associated type definitions. This eliminates the previous mechanism for instantiating OpenAI and Azure LLM clients within the indexing pipeline, indicating a shift to a different LLM integration strategy elsewhere in the codebase.

graphrag/index/llm · high confidence

Removal of legacy OpenAI LLM wrapper module

The \graphrag/query/llm/oai\ package has been removed, deleting the legacy OpenAI and Azure OpenAI wrapper implementations (including \ChatOpenAI\, \OpenAI\, \OpenAIEmbedding\, and their base classes). This change eliminates the internal code responsible for handling synchronous and asynchronous API calls, streaming responses, and retry logic for these specific providers, indicating a shift away from this implementation layer.

graphrag/query/llm/oai · high confidence

Removal of legacy Pydantic configuration models

The \graphrag/config/models\ directory has been removed, eliminating the legacy Pydantic-based configuration classes (such as \GraphRagConfig\, \LLMConfig\, \CacheConfig\, and \StorageConfig\) that previously defined the default configuration parameterization. This change removes the old configuration schema and its associated model definitions from the codebase.

graphrag/config/models · high confidence

Removal of legacy TypedDict-based configuration input models

The \graphrag/config/input\_models\ package has been removed, deleting all legacy configuration input classes (such as \GraphRagConfigInput\, \EntityExtractionConfigInput\, and \LLMParametersInput\) that were defined as \TypedDict\ structures. This change eliminates the previous configuration parameterization layer, meaning users can no longer import or rely on these specific model definitions for configuring GraphRAG components like chunking, entity extraction, or search.

_graphrag/config/input\models · high confidence

Removal of legacy community report extraction verbs

The \graphrag/index/verbs/graph/report\ module has been removed, eliminating the legacy \create\_community\_reports\ verb and its associated helper verbs (\prepare\_community\_reports\, \prepare\_community\_reports\_claims\, \prepare\_community\_reports\_edges\, \prepare\_community\_reports\_nodes\, and \restore\_community\_hierarchy\). This change also removes the \graph\_intelligence\ strategy implementation and its supporting type definitions, effectively deprecating the previous graph-based community report generation pipeline in favor of the current indexing engine architecture.

graphrag/index/verbs/graph/report · high confidence

Removal of legacy configuration module

The legacy configuration package located at graphrag/config has been completely removed. This deletion eliminates the previous implementation of configuration models, input models, environment readers, and default settings that were previously exported from the package root.

graphrag/config · high confidence

Removal of legacy context builder modules

The \graphrag/query/context\_builder\ package has been removed, deleting the legacy context-building infrastructure including base classes (\GlobalContextBuilder\, \LocalContextBuilder\), specific builders for community, entity, local, and source contexts, and the conversation history management utilities. This change eliminates the previous implementation of context assembly for system prompts, indicating a structural shift in how query context is generated within the application.

_graphrag/query/context\builder · high confidence

Removal of legacy graph clustering implementation

The legacy graph clustering verb and its associated Leiden strategy implementation have been removed from the indexing engine. This change eliminates the \cluster\_graph\ verb, the \GraphCommunityStrategyType\ enum, and the \strategies/leiden.py\ module that previously handled hierarchical graph clustering using the Leiden algorithm. Users relying on this specific clustering workflow will need to adopt the replacement implementation provided in the broader migration.

graphrag/index/verbs/graph/clustering · high confidence

Removal of legacy graph extraction and embedding modules

The graph indexing engine has removed the legacy implementation files for graph extraction, community reports, and node embeddings. Specifically, the \graphrag/index/graph/embedding\ module (Node2Vec embeddings), the \graphrag/index/graph/extractors\ package (including \ClaimExtractor\, \CommunityReportsExtractor\, \GraphExtractor\, and \SummarizeExtractor\), and the \graphrag/index/graph/visualization\ module (UMAP-based visualization) have been deleted. This change eliminates the old code paths for generating graph structures, extracting claims, and visualizing node positions, indicating a shift to a different underlying implementation for these graph operations.

graphrag/index/graph · high confidence

Removal of legacy graph query input retrieval utilities

The \graphrag/query/input/retrieval\ module has been completely removed. This deletes the \\_\init\\_.py\ file and all associated utility modules (\community\_reports.py\, \covariates.py\, \entities.py\, \relationships.py\, and \text\_units.py\) that previously provided functions for retrieving and formatting graph data (such as entities, relationships, and text units) into pandas DataFrames for query inputs.

graphrag/query/input/retrieval · high confidence

Removal of legacy graphrag index utility modules

The \graphrag/index/utils\ package has been completely removed, deleting all associated utility modules including \\_\init\\_.py\, \dataframes.py\, \ds\_util.py\, \json.py\, \load\_graph.py\, \tokens.py\, and \topological\_sort.py\. This eliminates legacy helper functions for DataFrame manipulation, JSON cleaning, graph loading, token counting, and topological sorting that were previously exposed via the utils namespace.

graphrag/index/utils · high confidence

Removal of legacy graphrag.index package and CLI

The entire \graphrag/index\ package has been removed, including the legacy CLI entry points (\\_\main\\_.py\, \cli.py\), the old pipeline configuration system (\create\_pipeline\_config.py\, \load\_pipeline\_config.py\), and the associated infrastructure for caching, storage, table emission, and progress reporting. This change eliminates the previous indexing engine implementation in favor of the new architecture.

graphrag/index · high confidence

Removal of legacy indexing verbs

The \graphrag/index/verbs\ module has been removed, eliminating a suite of legacy data processing verbs including \genid\, \embed\_graph\, \snapshot\, \spread\_json\, \text\_replace\, \text\_split\, \unzip\, and \zip\. This change reflects the migration of helper verbs and embedding operations to a new structure, meaning these specific indexing functions are no longer available in this location.

graphrag/index/verbs, graphrag/index/verbs/graph · high confidence

Removal of legacy mock LLM implementations

The \graphrag/llm/mock\ module has been removed, deleting the \MockChatLLM\ and \MockCompletionLLM\ classes. This change eliminates the previous mock LLM implementations that were used for testing, aligning with the broader replacement of the \graphrag.llm\ package with \fnllm\.

graphrag/llm/mock · high confidence

Removal of legacy pipeline configuration models

The legacy Pydantic-based configuration models for the indexing pipeline have been removed. This includes the deletion of \graphrag/index/config/\_\init\\_.py\ and all its submodules (\cache.py\, \input.py\, \pipeline.py\, \reporting.py\, \storage.py\, \workflow.py\), which previously defined types such as \PipelineConfig\, \PipelineCacheConfig\, \PipelineInputConfig\, \PipelineStorageConfig\, and \PipelineReportingConfig\. Users relying on these specific configuration classes for programmatic pipeline setup will need to migrate to the new configuration approach.

graphrag/index/config · high confidence

Removal of legacy prompt tuning templates

The prompt tuning module has removed its legacy prompt template files, including those for community report summarization, entity extraction, entity summarization, and example generation. Users relying on these specific prompt structures for graphrag prompt tuning will need to update their configurations or migrate to the new template system.

_graphrag/prompt\tune/template · high confidence

Removal of legacy query CLI and orchestration module

The \graphrag/query\ package has removed its legacy command-line interface and orchestration entry points. Specifically, the \\_\init\\.py\, \\\main\\_.py\, \cli.py\, \factories.py\, \indexer\_adapters.py\, and \progress.py\ files have been deleted. This eliminates the previous \argparse\-based CLI (invoked via \python -m graphrag.query\) and the associated factory methods for constructing local and global search engines, as well as the custom \StatusReporter\ and indexer data adapters that supported this older interface.

graphrag/query · high confidence

Removal of legacy question generation module

The \graphrag/query/question\_gen\ package, which previously provided the \BaseQuestionGen\ interface and \LocalQuestionGen\ implementation for generating follow-up questions based on conversation history and context, has been removed. This deletion eliminates the synchronous and asynchronous question generation capabilities, the associated \QuestionResult\ data structure, and the \QUESTION\_SYSTEM\_PROMPT\ that were previously available in this location.

_graphrag/query/question\gen · high confidence

Removal of legacy text chunking implementation

The legacy text chunking module in the indexing engine has been removed. This deletes the \text\_chunk\ verb and its associated strategy implementations (token-based and sentence-based), along with their type definitions. Users relying on this specific chunking verb for text processing will need to migrate to the updated chunking infrastructure.

graphrag/index/verbs/text/chunk · high confidence

Removal of legacy text embedding strategies and verb implementation

The \graphrag/index/verbs/text/embed\ module has been removed, deleting the \text\_embed\ verb and its associated strategy implementations (\openai\, \mock\) and type definitions. This eliminates the previous capability to generate text embeddings in-memory or store them in vector stores via the legacy configuration schema, indicating a shift to a different embedding mechanism elsewhere in the system.

graphrag/index/verbs/text/embed · high confidence

Removal of legacy text splitting module

The \graphrag/index/text\_splitting\ package has been removed from the codebase. This deletion eliminates the legacy text splitting logic, including the \TextSplitter\ base class, \TokenTextSplitter\, \TextListSplitter\, and the \check\_token\_limit\ utility, which are no longer part of the indexing engine's text processing pipeline.

_graphrag/index/text\splitting · high confidence

Removal of legacy v1 workflow definitions and loading infrastructure

The legacy v1 workflow system has been removed from the indexing engine. This change deletes the \default\_workflows.py\ file, which previously registered the standard v1 pipeline steps (such as \create\_base\_entity\_graph\, \create\_final\_communities\, and \create\_final\_text\_units\), along with the \load.py\ module responsible for instantiating and ordering these workflows. Users relying on the previous v1 workflow structure will no longer have these default definitions available in this package location.

graphrag/index/workflows · high confidence

Removal of legacy vector store implementations

The legacy vector store module has been removed, deleting the \\_\init\\_.py\ entry point and the specific implementations for Azure AI Search and LanceDB, along with their shared base classes (\BaseVectorStore\, \VectorStoreDocument\, \VectorStoreSearchResult\) and the \VectorStoreFactory\ registry. This eliminates the previous vector storage architecture from the codebase.

_graphrag/vector\stores · high confidence

Removal of prompt tuning configuration and input loaders

The \graphrag/prompt\tune/loader\ module has been removed, deleting the \config.py\, \input.py\, and \\\init\\_.py\ files. This eliminates the internal logic for reading GraphRAG settings from YAML/JSON files and loading document chunks for prompt tuning, indicating that this functionality is no longer supported or has been moved to a different location.

_graphrag/prompt\tune/loader · high confidence

Removal of prompt tuning generator module

The entire \graphrag/prompt\_tune/generator\ module has been removed. This eliminates the previous capability to automatically generate prompts for community summarization, entity extraction, entity summarization, and persona/domain inference via the LLM. Users relying on this automatic prompt tuning workflow will no longer have access to these specific generation functions.

_graphrag/prompt\tune/generator · high confidence

Removal of prompt tuning prompt definitions

The prompt definitions for the prompt tuning module (including those for community reporter role, domain, entity types, entity relationships, and persona generation) have been removed from the codebase. This change eliminates the hardcoded prompt templates previously located in the \graphrag/prompt\_tune/prompt\ directory, which will likely require users to rely on updated or external prompt sources for these specific tuning tasks.

_graphrag/prompt\tune/prompt · high confidence

Removal of structured search base classes and package init

The \graphrag/query/structured\search\ package has been removed, specifically deleting the \\\init\\_.py\ and \base.py\ files. This eliminates the \BaseSearch\ abstract class and the \SearchResult\ data class that previously defined the interface for local and global search implementations, along with their dependencies on context builders and LLM providers.

_graphrag/query/structured\search · high confidence

Removal of text\_translate verb and translation strategies

The text\_translate verb, along with its underlying OpenAI and mock translation strategies, has been removed from the indexing engine. Users can no longer use the text\_translate verb to translate text columns into different languages via LLMs or mock implementations within the indexing pipeline.

graphrag/index/verbs/text/translate · high confidence

Removal of v1 workflow definitions

The v1 workflow definitions in \graphrag/index/workflows/v1\ have been removed. This includes the deletion of files responsible for creating base and final entities, documents, text units, relationships, nodes, communities, covariates, and community reports, as well as helper workflows for joining text units to entity, relationship, and covariate IDs. Users relying on the v1 workflow structure will need to migrate to the updated workflow architecture.

graphrag/index/workflows/v1 · high confidence

Removed legacy graphrag LLM type definitions

The \graphrag/llm/types\ module has been deleted, removing the internal type definitions for LLM interactions (including \LLM\, \LLMCache\, \LLMConfig\, \LLMInput\, \LLMOutput\, and related callback and result types). This change aligns with the migration to the \fnllm\ library, meaning these specific protocol and data-class definitions are no longer part of the codebase.

graphrag/llm/types · high confidence

Behavioural changes

Automated dependency management and build script updates

This change introduces new automation scripts to manage workspace dependencies and GitHub operations, while updating existing build tooling. Specifically, it adds scripts to automatically close Dependabot pull requests, open issues for dependency sweeps, and update cross-package dependency versions within the workspace to match the current version. It also adds a script to copy build assets (like the LICENSE file) to package directories for PyPI distribution. Additionally, the semver-check script has been updated to use \uv run\ instead of \poetry run\, reflecting a shift in package management, and the legacy \e2e-test.sh\ script using Poetry has been removed.

scripts · high confidence

Migrate documentation site from Eleventy to MkDocs

The documentation site has switched its static site generator from Eleventy to MkDocs. This change removes the Eleventy configuration files (\.eleventy.js\, \.eleventyignore\, \.gitignore\) and the bundled Yarn binary, replacing them with a MkDocs-based build process. Users will now benefit from MkDocs' native support for static output directories and its integrated search capabilities, which are enabled for both local and global search within the documentation.

docsite · high confidence

Prompt tuning CLI refactored from argparse to Typer

The command-line interface for the prompt tuning feature has been migrated from the standard library \argparse\ to the \Typer\ framework. This change removes the legacy \\_\main\\_.py\ and \cli.py\ entry points, replacing them with a modern, structured CLI implementation that maintains the same functional capabilities for generating indexing prompts while improving command-line usability and code maintainability.

_graphrag/prompt\tune · high confidence

Removal of legacy LLM base classes and text utilities

The \graphrag/query/llm\ module has been cleaned up by removing the \\_\init\\_.py\, \base.py\, and \text\_utils.py\ files. This eliminates the previous abstract base classes for LLMs (\BaseLLM\, \BaseTextEmbedding\) and callbacks, as well as text processing utilities like \num\_tokens\ and \chunk\_text\ that relied on \tiktoken\. Users relying on these specific internal components within this directory will need to migrate to the new LLM provider implementations or utility functions provided elsewhere in the updated package structure.

graphrag/query/llm · high confidence

Removal of legacy covariates extraction implementation

The \extract\_covariates\ verb and its associated \graph\_intelligence\ strategy implementation have been removed from the indexing engine. This change eliminates the legacy code path for extracting claims (covariates) from text, including the \Covariate\ data models and the specific LLM-based extraction logic that relied on the \ClaimExtractor\. Users relying on this specific verb for claim extraction will need to migrate to the new helper verbs or updated covariate processing flows introduced in this release.

graphrag/index/verbs/covariates · high confidence

Test coverage

Added covariates test data for A Christmas Carol; Added integration tests for language model components; Added integration tests for storage backends and factory; Added integration tests for the CacheFactory; Added integration tests for vector store implementations; Added test infrastructure for GraphRAG LLM module; Added unit tests for BlobWorkflowLogger; Added unit tests for CSV and Parquet table providers; Added unit tests for LLM cache middleware event loop handling; Added unit tests for cache backends and bulk-read performance benchmarks; Added unit tests for configuration validation and loading; Added unit tests for core GraphRAG components; Added unit tests for graph algorithms; Added unit tests for indexing input loaders and utilities; Added unit tests for indexing operations; Added unit tests for indexing operations and updated config test imports; Added unit tests for prompt tuning document loading; Added unit tests for query context building, entity retrieval, and vector store filtering; Added unit tests for text encoding utilities; Added unit tests for the new cache configuration system; Removal of legacy LLM test helpers; Removal of unit tests for blob and file pipeline storage; Removed indexing config unit tests and fixtures; Removed legacy integration pipeline test suite; Removed unit tests for text split verb; Smoke test infrastructure modernization and stabilization; Updated notebook test infrastructure to use uv and target specific package notebooks; Updated unit tests for community report context sorting; Updated unit tests for graph intelligence entity extraction; Updated unit tests for stable\_lcc to use local graph definitions.

Dependencies

GraphRAG restructured into a multi-package monorepo with uv dependency management

The GraphRAG project has been reorganized from a single package into a monorepo containing eight distinct Python packages: graphrag, graphrag-cache, graphrag-chunking, graphrag-common, graphrag-input, graphrag-llm, graphrag-storage, and graphrag-vectors, all versioned at 3.2.0. This structural change is accompanied by a migration from Poetry to uv for dependency management, evidenced by the removal of the poetry.lock file and the addition of a \[tool.uv\] configuration in the root pyproject.toml. The unified-search-app package has also been updated to depend on graphrag 2.5.0, and the docsite's Eleventy-based build system has been removed.

(dependencies) · high confidence

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

This is the PUBLIC form of this artifact. Findings are listed in full, but the details of SECURITY findings — which rule fired, in which file, on which line, and how to fix it — are deliberately withheld, and any secret-scanner results are excluded entirely. Where detail is absent here it was REMOVED FOR PUBLICATION; it is not missing from the analysis. The complete artifact is available from the repository owner.

Score

  • CAI 62 → 63 (+0.5)
  • Rubric changed (rubric-2026.10.5 → rubric-2026.11.1) — scores are not directly comparable.

Lenses

  • Code Health 63 → 63 (+0.0)
  • Architecture 95 → 95 (+0.0)
  • Maturity 70 → 72 (+1.3)
  • Readiness 54 → 54 (+0.0)
  • Security 66 → 70 (+3.1)
  • Performance 100 → 100 (+0.0)

Resolved (14)

  • Documentation: no installation or build instructions (README.md)
  • Documentation: no installation or build instructions (docs/index.md)
  • Documentation: no usage examples (README.md)
  • Documentation: no usage examples (docs/index.md)
  • High CVE: [GHSA redacted] (unified-search-app/uv.lock)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • Off-boarding risk: anonymized user #1
  • Split packages/graphrag
  • Unpinned build actions

New (43)

  • Critical CVE: [GHSA redacted] (unified-search-app/uv.lock)
  • Critical CVE: [GHSA redacted] (uv.lock)
  • Critical vulnerability: [GHSA redacted] (unified-search-app/uv.lock)
  • Critical vulnerability: [GHSA redacted] (uv.lock)
  • High CVE: [GHSA redacted] (uv.lock)
  • High CVE: [GHSA redacted] (uv.lock)
  • High CVE: [GHSA redacted] (uv.lock)
  • High: security finding (details withheld)
  • High: security finding (details withheld)
  • Low vulnerability: [GHSA redacted] (uv.lock)
  • Off-boarding risk: anonymized user #1
  • Orphaned file (packages/graphrag-storage/graphrag_storage/tables/table_provider_factory.py)
  • Orphaned file (packages/graphrag-vectors/graphrag_vectors/cosmosdb.py)
  • Orphaned file (packages/graphrag-vectors/graphrag_vectors/filtering.py)
  • Orphaned file (packages/graphrag-vectors/graphrag_vectors/timestamp.py)
  • Orphaned file (packages/graphrag/graphrag/config/models/extract_graph_nlp_config.py)
  • Orphaned file (packages/graphrag/graphrag/data_model/data_reader.py)
  • Orphaned file (packages/graphrag/graphrag/data_model/dfs.py)
  • Orphaned file (packages/graphrag/graphrag/data_model/schemas.py)
  • Orphaned file (packages/graphrag/graphrag/graphs/connected_components.py)
  • …and 23 more

Changes since last survey

  • 2 commits — 2 feature/other, 0 fixes

By area

  • .github/workflows — 1 commit
  • scripts/open-deps-update-issue.ts — 1 commit

Notable commits

  • change: Pin maintenance workflow actions to full commit SHAs (#2615)
  • change: Update dep sweep script.

Architecture

  • Unchanged — 0 containers · 1 contexts · 0 edges

Written by watchdog.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

microsoft/graphrag was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 3 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 34b2b67ae4601ad0b47bcd179fbe3f2a23aa0857 — the exact code this score is about.
  • Scored under rubric-2026.11.1 — the same rubric and the same method as every other entry in this index.
  • Measured by watchdog.canine.dev using codehealth-analyzer measure-2026.11.1+preprod-7b4ed5af9e84.