RAG chatbot on Bedrock
Estimate a retrieval-augmented chatbot: OpenSearch vectors + Bedrock generation + Lambda.
Use this for a 'chat over our documents' product with a real vector store. Includes one-time embedding ingestion plus the ongoing OpenSearch Serverless floor.
Estimate your workload
Model choice is usually the biggest cost lever. Nova Micro/Lite are ~10–70× cheaper than Premier.
OpenSearch Serverless has a minimum always-on OCU floor.
Results update automatically as you type.
- OpenSearch Serverless(2 OCU + 20 GB)
- $350.88
- Bedrock generation (Amazon Nova Pro)(per-query tokens)
- $241.92
- Lambda (orchestration)
- $0.4120
- API Gateway
- $0.0600
- Query embeddings
- $0.0480
- Per month(all services)
- $593.32
- Per year
- $7,119.84
- Embedding ingestion (one-time)(30,000 chunks)
- $0.4800
| OpenSearch Serverless | $350.88 | 59% |
| Bedrock generation (Amazon Nova Pro) | $241.92 | 41% |
| Lambda (orchestration) | $0.4120 | 0% |
| API Gateway | $0.0600 | 0% |
| Query embeddings | $0.0480 | 0% |
Frequently asked questions
Why is there a fixed monthly cost even with no traffic?+
OpenSearch Serverless bills a minimum always-on OCU floor whether or not you query it. That floor often dominates the bill for small RAG apps.
What is the one-time ingestion cost?+
Every document is chunked and embedded once when indexed. That embedding pass is a one-time charge, shown separately from your recurring monthly cost.
Other AWS architectures
AgentCore and managed-service rates last verified 14 July 2026 (us-east-1). Bedrock model prices last reviewed 28 July 2026. Estimates for planning, not a quote.