spring-ai-simple-chat-client

Spring AI Simple Chat Client

A lightweight Spring Boot application demonstrating Spring AI with Amazon Bedrock without the AgentCore starter. This shows traditional REST endpoints for comparison.

Features

• Traditional REST API: Manual endpoint creation (not using @AgentCoreInvocation)
• Synchronous and Streaming: Both response types supported
• Custom Tool Integration: Date/time tools available to AI
• Amazon Bedrock: Integration with Claude 3 Sonnet model

Comparison with AgentCore Examples

This example uses manual REST controllers instead of the AgentCore starter:

• Manual @PostMapping endpoints vs @AgentCoreInvocation
• Custom paths (/ai, /ai/stream) vs fixed /invocations
• No automatic AgentCore health monitoring

Prerequisites

• Java 21
• Maven
• AWS account with access to Amazon Bedrock
• AWS credentials configured locally

Configuration

Uses Amazon Bedrock's Claude 3 Sonnet model in EU West 1:

spring.application.name=agents
spring.ai.bedrock.aws.region=eu-west-1
spring.ai.bedrock.converse.chat.options.model=eu.anthropic.claude-3-7-sonnet-20250219-v1:0

Building and Running

mvn spring-boot:run

The application starts on port 8080.

API Endpoints

Synchronous Chat

curl -XPOST 'http://localhost:8080/ai' \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Tell me about Spring AI in 2 sentences."}'

Streaming Chat

curl -XPOST -N 'http://localhost:8080/ai/stream' \
  -H "Content-Type: application/json" \
  -d '{"prompt":"Who is George Mallory?"}'

With Tool Use

curl -XPOST -N 'http://localhost:8080/ai/stream' \
  -H "Content-Type: application/json" \
  -d '{"prompt":"What is the current date and time?"}'

Project Structure

• ChatController.java: Manual REST endpoints (no AgentCore)
• PromptRequest.java: Request payload record
• DateTimeTools.java: Custom tool for date/time information
• AgentsApplication.java: Spring Boot application entry point

Key Differences from AgentCore Examples

1. Manual Endpoints: Uses @PostMapping instead of @AgentCoreInvocation
2. Custom Paths: /ai and /ai/stream instead of /invocations
3. No Health Monitoring: No automatic /ping endpoint
4. More Code: Requires manual controller setup vs single annotation