Skip to main content

Workflow Change Log 1.3

24/09/2026

GraphRAG 1.3

The GraphRAG Workflow Change Log provides an overview of changes made to the workflows specifying the affected workflows, change type and whether the change is breaking or not. In the last column Migration notes you will find further relevant details.

Change

Affected workflows

Type

Breaking change?

Migration notes

Hidden labels and concept scheme details of concepts were added

Configuration

Fetch Concept Details

Bug fix

No

skos:hiddenLabel and skos:inScheme are now enabled by default as conceptProperties in the Configuration workflow

The Fetch Concept Details workflow is now usable as a standalone webhook

Fetch Concept Details

Step - Graph Modeling Concept Enricher - Sync

Improvement

No

-

Various configuration variables and workflows were renamed for clarity

Configuration

Error - Async, Error - Sync

Main

Parallel - Main, Parallel - Steps Batching

Parallel Post Final Answer - Main

Step - Final Answer - Sync

Step - GraphDB MCP - Async

Step - Input guardrails - Main

Step - Vector Search - Async

Utils - Calculate LLM usage

Vector search tool

Bug fix

No

Typos in configuration variables were fixed (e.g. parallelStepMaxTiemout -> parallelStepMaxTimeout, vectorStoreCustomMappins -> vectorStoreCustomMappings)

Explainability messages with MCP and vector search executionIds were added for testing and evaluation purposes

Parallel - Main, Step - GraphDB MCP - Async

Step - Vector Search - Async

Improvement

No

-

Custom concept properties to be added to the LLM context can now be defined

Configuration

Fetch Concept Details

Improvement

No

A new configuration variable conceptPropertyLlmContext was introduced in the Configuration workflow to be able to add any custom concept properties to the LLM context

The authentication method in all Graph Modeling requests is now OAuth2 replacing Basic Auth

Fetch Concept Details

Step - Graph Modeling Concept Enricher - Sync

Step - Graph Modeling Expansion - Async

Credential Graph Modeling

Improvement

Yes

The Graph Modeling user's credential were replaced with Oauth2 credentials which require setting up a GraphRAG client in the Keycloak instance used by Graph Modeling (refer to the Custom Keycloak Client section for more details on how to set up the GraphRAG client in Keycloak)Custom Keycloak Client

The n8n API key data table was converted to a credential

Error - Async

Error - Sync

Utils - Calculate LLM usage

Improvement

Yes

Improved structured output parsing reliability across workflows by adding fallback JSON handling for models that struggle with schema enforcement (for example, gpt-5.6-terra)

Configuration, LLM call

Step - Final Answer - Sync

Improvement

No

The MCP protocol for connecting to the GraphDB MCP server was changed from SSE to HTTP Streamable

Step - GraphDB MCP - Async

Improvement

Yes

The GraphDB instance used for GraphRAG no longer requires the flag graphdb.mcp.transport.mode=sse. The graphDBMcpUrl is by default "http://graphdb-host:port/mcp" In the Configuration workflow.