GraphRAG Configuration Guide - Infrastructure & Payload Optimization (Helm Layer)

30/06/2026

GraphRAG version 1.2.0

Apply the following modifications within your Helm chart values.yaml environment blocks when configuring infrastructure resources:

AWS Cloud Constraints
When routing to AWS Bedrock embeddings or AWS OpenSearch clusters, all required AWS authentication variable keys must be injected explicitly via Kubernetes Secrets or Helm container environment properties maps.
Network Performance Tuning
For high-throughput production environments, set INCLUDE_VECTOR_DATA: "false" inside your components service environment property block. This explicitly instructs the application to discard bulky raw numerical vector arrays during message transport, drastically lowering cross-pod network latency.

Migrating to a different vector backend, such as switching from OpenSearch to an external Qdrant instance, requires a coordinated two-layer update. Follow these steps to transition without disrupting the pipeline:

Because the Components Service establishes the physical connection to the vector database, its underlying deployment properties must be updated first.

Open your custom Helm chart values.yaml template file.
Update the target parameters block (such as connection URLs, internal endpoint routes, and resource mappings) to reflect the new vector database server.
Apply the infrastructure changes and trigger a rolling update of the components microservice pods by running:
helm upgrade <release-name> charts/graphrag/

Once the physical service layer is reconnected, update the application orchestrator's payload transformation rules dynamically.

Access the Workflows dashboard interface and open the Configuration sub-workflow.
Open the main Configuration Code JavaScript node.
Locate the vectorStorePreset parameter string and switch its value to your new target provider pattern (e.g., change from "opensearch_native" to "qdrant").
Click Save/Publish to commit your change.

In this section: