Kendra GenAI Index vs Vector Store Price for Knowledge Base Creation and Use

When comparing pricing for knowledge base options in AWS Bedrock, there are important differences between using Amazon Kendra GenAI Index and a vector store approach.

For AWS Bedrock knowledge bases, there is no specific charge for the knowledge base feature itself. However, you will incur costs for two main components:

1. The AI models you use (such as embedding models like Embed English v3)
2. The vector database that stores your embeddings

If you choose the "Quick create a new vector store" option with Bedrock knowledge bases, this typically provisions an Amazon OpenSearch Serverless instance. The costs will depend on the storage and compute resources consumed by your vector database.

For Amazon Kendra GenAI Index, the pricing structure is more defined:

• Base index capacity: $0.32 per hour
• Storage units: $0.25 per hour per unit (each unit supports approximately 22,222 documents)
• Query units: $0.07 per hour per unit (each unit supports approximately 0.15 queries per second)

For example, if you have 200,000 documents and need to support about 25,000 searches per day (roughly 0.28 queries per second), you would need a base Kendra GenAI Index, 9 storage capacity units, and 2 additional query capacity units.

The Kendra GenAI Enterprise Edition offers the highest accuracy for retrieval operations and RAG use cases. It's powered by advanced information retrieval technologies like hybrid search (keyword and vector), semantic embedding, and re-ranker models. It also enables mobility of your indexed data across AWS generative AI services, allowing integration with Amazon Bedrock tools like agents and prompt flows.

When deciding between these options, consider not just the pricing but also your specific use case requirements, the volume of documents, expected query patterns, and the need for advanced search capabilities.
Sources
Pricing for AWS Bedrock knowledge bases | AWS re:Post
Amazon Kendra Pricing - Amazon Web Services
Cost - Generative AI Application Builder on AWS
Index types in Amazon Kendra - Amazon Kendra