Components Service - Environment Configuration

15/10/2026

GraphRAG 1.3.0

This article outlines specifics of Components service environment configuration and relevant configuration parameters of GrapRAG workflows.

As the Components service is a separate Spring service running inside a Docker container (Kubernetes pod), it maintains its own set of environment variables that are required at startup. Any changes to these environment variables require a service restart or redeployment to take effect.

EMBEDDING_PROVIDER:
Specifies the embedding service provider. Accepted values are aws or openai.
Credentials for the embedding provider:
The required credentials differ depending on the selected provider:
- If EMBEDDING_PROVIDER=aws:
  - AWS_REGION : Optional. Specifies the AWS region. Default is us-east-1.
  - AWS_ACCESS_KEY_ID : Required if the embedding provider is AWS or if OpenSearch is hosted on AWS.
  - AWS_SECRET_ACCESS_KEY: Required if the embedding provider is AWS or if OpenSearch is hosted on AWS.
- If EMBEDDING_PROVIDER=openai:
  - OPENAI_API_KEY: Required. The API key for OpenAI.
  - OPENAI_BASE_URL: Optional. The URL where the OpenAI embeddings service is hosted. Defaults to https://api.openai.com/v1/embeddings.
VECTOR_STORE:
Specifies the vector store type. Accepted values are opensearch, opensearch_selfhosted, or elasticsearch. Each store type uses a different method for constructing search queries.
VECTOR_INDEX:
The index against which similarity queries will be executed.
Vector store address:
Use ELASTIC_URL for Elasticsearch or OPENSEARCH_ENDPOINT for OpenSearch.
Vector store credentials:
Required if the vector store is secured. Currently, only secured Elasticsearch is supported.

The following configurations can be modified after the service is already running. They are located in the Configuration GraphRAG Workflows workflow:

graphRagComponentsUrl (required):
The internal service name for Docker or Kubernetes. For example: http://graphrag-components:8080.
vectorSearchUrl (required):
The address of the vector search endpoint. Typically set to ${graphRagComponentsUrl}/vector/search.
vectorSearchTopK (optional, default = 5):
The number of documents returned.
vectorStorePreset (optional, default = custom):
The mappings between the raw vector store field names in the returned data and the desired field names in the final output from the vector search workflow or tool. Supported values are: bedrock_knowledgebase, opensearch_native, pinecone, qdrant, weaviate, and custom. Use custom when the field names in your vector store differ from the default names.
vectorStoreCustomMappings (required if vectorStorePreset = custom):
The JSON paths from the vector store response must be mapped to the following field names:
- payload: The text that was embedded.
- source: The URL of the document.
- score: The similarity score relative to the search query or text.
- id: The unique identifier of the document
  Example value
  "vectorStoreCustomMappings": { "payload": "_source.content", "source": "_source.metadata.source_uri", "score": "_score", "id": "_id" }
- vectorStoreMetadataExclude (optional):
  An array of properties to exclude (set to null) from the passed-through metadata (or properties, payload, or source equivalent). For example: ["source_link", "content"].
- vectorStoreMetadataPassthrough (optional, default = true):
  Specifies whether to pass through the metadata from the vector store results.
- vectorIndex (required only for Pinecone, Qdrant or Weaviate):
  The index name for the vector store.
- embeddingsProvider (required only for Pinecone, Qdrant or Weaviate):
  The embeddings provider. Supported values: aws, openai.
  Note
  The Components service is not required for Pinecone, Qdrant and Weaviate. Dedicated GraphRAG workflow nodes are available for these vector stores.

In this section:

Components Service - Environment Configuration

Note