OCI GenAI Cloud Operations Agent

This repository provides the necessary Terraform configuration and container deployment artifacts to deploy an AI-powered Cloud Operations Assistant on Oracle Cloud Infrastructure (OCI). The solution combines OCI Generative AI Agents, Model Context Protocol (MCP) tooling, OCI Logging, Streaming, and Retrieval-Augmented Generation (RAG) capabilities to create an intelligent operational assistant for cloud infrastructure management and troubleshooting.

The solution deploys:

• OCI Generative AI Agent
• Knowledge Base with RAG integration
• MCP Server for OCI operational tooling
• Application Server UI
• OCI Logging and Streaming resources
• Test compute infrastructure
• OCI Container Registry repositories

How It Works

The repository is structured as a Terraform-based deployment project combined with containerized application services.

• infra-stack/:
  • kb_file: Sample process document for handling high cpu usage alerts.
  • modules: Reusable Terraform modules.
  • CloudOps_Infra_RMS.zip: Complete Resource Manager stack to deploy solution infrastructure components.
  • main.tf: Main Terraform file that sets up the infrastructure components.
  • output.tf: Output Terraform file to record output variables for other RMS stack.
  • schema.yaml: OCI Resource Manager schema for guided deployment.
  • variable.tf: Configure the OCI variables for the infrastructure components.
• container_images/:
  • MCP_Server/:
    • app.py: Application file for the MCP server.
    • Dockerfile: Main docker file to create the application container image:
    • README.md: Sample readme file for isolated instance deployment.
    • requirements.txt: Requirements file for python packages needed for the deployment.
  • Agent and UI/:
    • app.py: Application file for the frontend server.
    • Dockerfile: Main docker file to create the application container image:
    • README.md: Sample readme file for isolated instance deployment.
    • requirements.txt: Requirements file for python packages needed for the deployment.
  • README.md: Instructions file to create and store the container images in OCIR created in last deployment step.
• container-instances-stack/:
  • modules: Reusable Terraform modules.
  • Container_Instance_RMS.zip: Complete Resource Manager stack to deploy MCP and frontend application container instances.
  • main.tf: Main Terraform file that sets up the container instances.
  • schema.yaml: OCI Resource Manager schema for guided deployment.
  • variable.tf: Configure the OCI variables for the infrastructure components.
  • version.tf: Terraform file that controls the TF provider version for OCI.

The solution works as follows:

1. A user interacts with the Application Server UI.
2. Requests are sent to the OCI Generative AI Agent.
3. The agent determines whether operational tooling is required.
4. If needed, the agent invokes MCP tools exposed through the MCP Server.
5. The MCP Server interacts with OCI services using OCI SDKs and APIs.
6. Results are returned back to the AI Agent.
7. The AI Agent generates a contextual response for the user.

Solution Deployment

The solution is deployed in three steps.

1. Deploy the Infrastructure Stack
2. Build and Push the MCP Server and Application container images.
3. Deploy the Container Instance Stack

Part 1: Deploy the Infrastructure Stack

Note: The following steps deploy the base OCI infrastructure, AI Agent resources, logging pipeline, and container registry repositories required for the solution.

1. Clone the repository from GitHub.
2. Use Oracle Resource Manager to create and apply the stack.
   • using the hamburger menu, go to Oracle Resource Manager
   • choose Stacks
   • click Create stack
   • select My configuration
   • in the configuration section select zip
   • upload the cloudops-infra-rms.zip from the repository
   • provide a meaningful stack name
   • click Next
   • choose the target compartment
   • provide the required variables:
     • VCN and subnet information
     • SSH public key
     • compartment OCIDs
     • region configuration
   • click Next
   • select Run apply
   • click Create
3. Wait for the stack deployment to complete successfully.
4. After deployment, collect the outputs from the stack:
   • MCP OCIR repository path
   • Application OCIR repository path
   • Agent Endpoint
   • Stream OCIDs

Part 2: Create and Push the OCIR Images

The solution uses two container images:

• MCP Server image
• Application Server image

Prerequisites

• Docker installed
• Access to OCI Container Registry (OCIR)
• Auth token for OCI registry login

Build and Push Images

1. Login to OCIR

   docker login <region-key>.ocir.io
   

   Example:

   docker login iad.ocir.io
   

   Use:

   • OCI Username
   • OCI auth token

2. Build and Push MCP Server Image

   1. Navigate to the MCP Server directory:

      cd mcp-server
      

   2. Build and tag the mcp server image

      docker build . -t <mcp-repository-path>:latest
      

   3. Push the mcp server image

      docker push <mcp-repository-path>:latest
      

3. Build and Push Application Server Image

   1. Navigate to the Application Server directory:

      cd agent-and-ui
      

   2. Build and tag the mcp server image

      docker build . -t 'application-server-repository-path':latest
      

   3. Push the application server image

      docker push 'application-server-repository-path':latest
      

Part 3: Deploy the Application and MCP Containers

Once the container images are available in OCIR, deploy the runtime components.

Deployment Requirements

You will need:

• Agent Endpoint from Part 1
• OCIR image paths from Part 2
• OCI subnet configuration

Recommended Deployment Topology

• MCP Server
  • Deploy in:
    • Private subnet
• Application Server
  • Deploy in:
    • Public subnet
  • Expose:
    • TCP Port 8080

Deploy with OCI Resource Manager

1. Use Oracle Resource Manager to create and apply the stack.
   • using the hamburger menu, go to Oracle Resource Manager
   • choose Stacks
   • click Create stack
   • select My configuration
   • in the configuration section select zip
   • upload the container-instance-rms.zip from the repository
   • provide a meaningful stack name
   • click Next
   • choose the target compartment
   • provide the required variables:
     • Display Name Prefix
     • Availability Domain
     • MCP Container Image URL from last step
     • Application Container Image URL from last step
     • RAG Agent Endpoint
     • Container Configuration
       • Shape
       • OCPU
       • Memory
     • VCN and subnet information
     • Regional GenAI Endpoint
     • GenAI Model ID
   • click Next
   • select Run apply
   • click Create
2. Wait for the stack deployment to complete successfully.
3. After deployment, validate the container information
   1. View Application container instance IP address
   2. Open a new browser window to the following address
      1. http://application server public IP:8080

Post-Installation

Once the deployment is complete, the AI-powered DevOps Agent can be used for:

• OCI operational troubleshooting
• Infrastructure visibility
• Monitoring and alarm inspection
• Log analysis
• AI-assisted DevOps workflows
• Natural language operational automation

Clean-Up

1. Navigate to Oracle Resource Manager
2. Select the deployed stack
3. Click Destroy