- Graphwise Platform Documentation
- GraphRAG
- Components Service
- Supported Embedding Models & Vector Stores
Supported Embedding Models & Vector Stores
15/10/2026
GraphRAG 1.3.0
All Titan and Cohere models are supported for Amazon Bedrock. For more details please refer to Amazon Titan respectively Cohere documentation. Supported models include:
amazon.titan-embed-text-v2:0cohere.embed-english-v3
All OpenAI embedding models are supported. For more details please refer to the OpenAI documentation.
Note
Certain embedding models may not be available in all regions or subscription tiers. Please consult your AWS or OpenAI dashboard to verify which models are accessible to you.
Caution
The embedding model used to ingest documents into the vector store must be identical to the model used in the Components service. A mismatch will result in either errors due to incompatible vector dimensions or the retrieval of irrelevant results.
The supported stores are Elasticsearch, OpenSearch and AWS OpenSearch Serverless. Both Elasticsearch and OpenSearch natively support vector capabilities without requiring external third-party plugins; however, they approach initial index configuration differently.
Caution
If an unsupported vector store is specified in the VECTOR_STORE environment variable, the service will not start. The following error message will appear in the logs: "Unsupported vector store: <the specified store>"