Skip to content

New serverless pattern - apigw-lambda-websearch-bedrock-cdk#3183

Open
NithinChandranR-AWS wants to merge 1 commit into
aws-samples:mainfrom
NithinChandranR-AWS:NithinChandranR-AWS-feature-apigw-lambda-websearch-bedrock-cdk
Open

New serverless pattern - apigw-lambda-websearch-bedrock-cdk#3183
NithinChandranR-AWS wants to merge 1 commit into
aws-samples:mainfrom
NithinChandranR-AWS:NithinChandranR-AWS-feature-apigw-lambda-websearch-bedrock-cdk

Conversation

@NithinChandranR-AWS

Copy link
Copy Markdown
Contributor

Description

Grounded AI answers using Amazon Bedrock AgentCore Web Search + Amazon Bedrock. Deploys an Amazon API Gateway REST API backed by an AWS Lambda function that searches the live web for current facts, then uses Amazon Bedrock to synthesize accurate, cited answers.

Architecture

Amazon API Gateway → AWS Lambda → Amazon Bedrock AgentCore Web Search (MCP/SigV4) → Amazon Bedrock Claude (inference with citations) → grounded response with source URLs

What makes this different

This is RAG with the live web as the knowledge base — no vector database, no embedding pipeline, no data ingestion. The web is always current.

The Lambda orchestrates a genuine two-step workflow:

  1. Web search via AgentCore Gateway (MCP protocol, SigV4 signed)
  2. Bedrock inference with search results as context (citation-aware prompt)

Neither step can replace the other. The gateway connector provides facts; Bedrock provides reasoning and citation formatting.

Deployed and tested

  • Stack deploys successfully in us-east-1 (17 resources)
  • API Gateway endpoint returns grounded answers with 5 cited sources
  • Direct Lambda invocation also works
  • Citations reference real URLs from web search results

…h + Bedrock

Deploy Amazon API Gateway + AWS Lambda that orchestrates Amazon Bedrock
AgentCore Web Search for live web facts, then Amazon Bedrock for
inference to produce cited answers grounded in current information.

Architecture: API Gateway -> Lambda -> AgentCore Web Search (MCP/SigV4)
-> Bedrock Claude (RAG with live web as knowledge base) -> cited response.

Eliminates need for vector databases or embedding pipelines — the web
IS the knowledge base, always current.
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