confluentinc/quickstart-streaming-agents

This repository is a hands-on quickstart for building AI agents that run inside Confluent Cloud, a managed cloud platform built on Apache Kafka and Apache Flink. Kafka is a high-throughput event streaming system (think of it as a very fast, scalable message bus), and Flink is a stream-processing engine that can run computations on data as it flows through in real time. This quickstart shows how to combine those tools with AI to build agents that react to live data streams rather than waiting for batch jobs. The repo contains four labs. Lab 1 builds a price-matching agent that scrapes competitor websites and adjusts product prices in real time using MCP tool calling (a protocol for connecting AI models to external tools). Lab 2 demonstrates vector search and RAG, retrieval-augmented generation, where you pre-load documents into a searchable database so an AI can look up relevant information before answering. Lab 3 builds an end-to-end boat fleet management system combining agent definitions, tool calling, vector search, and anomaly detection. Lab 4 builds a fraud detection system for insurance claims that identifies suspicious patterns in real time using anomaly detection and AI analysis. The target users are developers or data engineers who want to see how AI agents can be embedded directly into event-driven data pipelines on Confluent Cloud, using AWS Bedrock or Azure OpenAI as the LLM provider. Setup uses Terraform for infrastructure provisioning and a Python deployment script. Requires a Confluent Cloud account.