Skip to content

New serverless pattern - agentcore-opensearch-serverless-nextgen-cdk#3132

Open
NithinChandranR-AWS wants to merge 1 commit into
aws-samples:mainfrom
NithinChandranR-AWS:NithinChandranR-AWS-feature-agentcore-opensearch-serverless-nextgen-cdk
Open

New serverless pattern - agentcore-opensearch-serverless-nextgen-cdk#3132
NithinChandranR-AWS wants to merge 1 commit into
aws-samples:mainfrom
NithinChandranR-AWS:NithinChandranR-AWS-feature-agentcore-opensearch-serverless-nextgen-cdk

Conversation

@NithinChandranR-AWS

@NithinChandranR-AWS NithinChandranR-AWS commented Jun 10, 2026

Copy link
Copy Markdown
Contributor

Description

Deploys an AI agent on Amazon Bedrock AgentCore Runtime that uses an Amazon OpenSearch Serverless NextGen collection as a search tool. The agent receives natural language queries and searches the OpenSearch collection using the opensearch-py SDK.

Architecture

User -> AgentCore Runtime (Strands Agent) -> OpenSearch Serverless NextGen (SEARCH collection)

Key Features

  • First AgentCore + OpenSearch Serverless combo pattern in this repo
  • OpenSearch Serverless NextGen collection group (scale-to-zero)
  • AgentCore Runtime with ARM64 container (Python Strands agent)
  • IAM role with inline policies (atomic permissions for AgentCore validation)
  • Encryption, network, and data access policies for OpenSearch Serverless
  • `@tool` decorator for natural language search via opensearch-py

Framework / Language

  • AWS CDK (TypeScript)
  • Agent: Python 3.12 (Strands Agents + opensearch-py)

Testing

  • CDK synth verified (9 CloudFormation resources)
  • Deployed to AWS: AgentCore Runtime READY, OpenSearch collection ACTIVE
  • ARM64 image built via AWS CodeBuild and pushed to ECR
  • IAM: least-privilege with inline policies for atomic attachment"

…t with OpenSearch Serverless NextGen search tool
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants