Skip to main content

Command Palette

Search for a command to run...

Supercharging Microservices with an Intelligent Multi-Agent System on GKE

Updated
3 min read
Supercharging Microservices with an Intelligent Multi-Agent System on GKE

The world of microservices is all about agility and scalability. But what if we could make them smarter? What if we could add a layer of intelligence that understands user needs and proactively provides assistance? That's exactly what we set out to do for the GKE Turns 10 Hackathon.

We built an Intelligent Multi-Agent System on Google Kubernetes Engine (GKE) that supercharges existing microservice applications, Bank of Anthos and Online Boutique, with a powerful AI brain. The best part? We did it without touching a single line of the original application code.

The Challenge: AI-Powered Microservices

Our goal was to enhance the user experience of these applications by providing a unified, intelligent assistant that could help users with their banking and shopping needs. We wanted to create a system that could:

  • Understand user intent: Whether a user wants to check their balance, find a product, or get financial advice, the system should understand their needs.

  • Provide holistic guidance: By combining information from both the banking and shopping applications, the system can offer comprehensive financial advice.

  • Be proactive: The system should be able to anticipate user needs and offer suggestions before they even ask.

The Solution: A Multi-Agent System on GKE

To achieve this, we designed a multi-agent system that runs on GKE. Each agent is a specialized AI model responsible for a specific task:

  • Banking Agent: Handles all interactions with the Bank of Anthos application, such as checking balances, transferring funds, and providing transaction history.

  • Shopping Agent: Interacts with the Online Boutique application to search for products, make recommendations, and manage the shopping cart.

  • Financial Wellness Agent: Provides personalized financial advice based on the user's spending habits and financial goals.

  • Predictive Analytics Agent: Uses historical data to predict future financial trends and provide proactive recommendations.

  • Infrastructure Agent: Monitors the health and performance of the GKE cluster and the microservices.

  • Unified Intelligence Orchestrator: The "brain" of the system, responsible for understanding user requests and routing them to the appropriate agent. It uses Google's Gemini model to understand natural language and orchestrate the conversation flow.

Architecture

Our architecture is designed to be scalable, resilient, and secure. Here's a high-level overview:

  • Frontend: A Vue.js web interface with a real-time chat component that allows users to interact with the intelligent agents.

  • Backend: A Flask-based backend that hosts the Unified Intelligence Orchestrator and the other agents.

  • Infrastructure: The entire system is deployed on a GKE cluster, which provides autoscaling, self-healing, and other managed Kubernetes features.

  • Integration: The agents interact with the Bank of Anthos and Online Boutique microservices through their existing APIs.

Key Technologies

We used a variety of Google Cloud technologies to build our solution:

  • Google Kubernetes Engine (GKE): For deploying and managing our containerized applications.

  • Google Gemini: To power the natural language understanding and generation capabilities of our agents.

  • Flask: A lightweight Python web framework for building the backend.

  • Vue.js: A progressive JavaScript framework for building the frontend.

What's Next?

We're just scratching the surface of what's possible with intelligent multi-agent systems on GKE. In the future, we plan to:

  • Add more agents: We want to expand the capabilities of our system by adding new agents for tasks like travel planning, bill payment, and more.

  • Improve personalization: We want to use machine learning to provide more personalized recommendations and advice.

  • Integrate with more applications: We want to connect our system to other third-party applications and services.

We're excited about the potential of this technology to revolutionize the way we interact with software. By combining the power of AI with the scalability of GKE, we can build intelligent systems that are truly helpful and proactive.


This blog post was created for the GKE Turns 10 Hackathon.