AI Teammates

Use terraform to manage Kubiya AI Teammates

AI Teammates are the heart of Kubiya. As an operator, you may prefer to create and manage your AI Teammates using IaC.

Want to know more about AI Teammates in Kubiya? Learn more

Note: Creating an AI Teammate requires having already created a runner and a source, as these are mandatory fields for AI Teammates.

Here's an example of how to provision an AI Teammate using Terraform:

terraform {
  required_providers {
    kubiya = {
      source  = "kubiya-terraform/kubiya"
    }
  }
}

provider "kubiya" {
   //Your Kubiya API Key will be taken from the 
  //environment variable KUBIYA_API_KEY
  //To set the key, please use export KUBIYA_API_KEY="YOUR_API_KEY"
}

resource "kubiya_agent" "agent" {

  //Mandatory Fields
  name            = "Kubernetes Admin"             //String
  runner          = "runner-dev-cluster"           //String
  description     = "This agent can perform various Kubernetes tasks using kubectl"     
  instructions = "You are an AI agent specialized in managing Kubernetes clusters using kubectl. Your tasks include monitoring pod statuses, scaling deployments, updating container images, and reporting issues to JIRA."       
  
  //Optional fields, String
  model = "azure/gpt-4"  // If not provided, Defaults to "azure/gpt-4"
  //If not provided, Defaults to "ghcr.io/kubiyabot/kubiya-agent:stable"
  image = "ghcr.io/kubiyabot/kubiya-agent:stable" 
  
  //Optional Fields (omitting will retain the current values): 
  //Arrays
  secrets = []
  integrations = ["github","jira","aws"]
  users = ["[e-mail]"]                     
  groups = ["Admin"]                                 
  links = []
                                     
  //Objects
  starters = [
    {
      name = "Check pod status"
      command      = "kubectl get pods"
    },
    {
      name = "Scale deployment"
      command      = "kubectl scale deployment my-deployment --replicas=3"
    }
  ]
  
  tasks = [
    {
      name        = "Check Container Image Backoff"
      prompt      = "Check all namespaces for pods in ImagePullBackOff state using 'kubectl get pods --all-namespaces'. If found, report them to JIRA."
      description = "Use kubectl to find pods in ImagePullBackOff state and report them to JIRA."
    },
    {
      name        = "Update Container Image"
      prompt      = "Ask the user for the deployment name and the new image. If a specific namespace is provided, filter deployments by that namespace using 'kubectl get deployments -n <namespace>'. Then, update the deployment's container image using 'kubectl set image deployment/<deployment-name> <container-name>=<new-image>'."
      description = "Use kubectl to update the container image for the specified deployment."
    },
    {
      name        = "Scale Deployment"
      prompt      = "Ask the user for the deployment name and the desired number of replicas. If a specific namespace is provided, filter deployments by that namespace using 'kubectl get deployments -n <namespace>'. Then, scale the deployment using 'kubectl scale deployment/<deployment-name> --replicas=<number-of-replicas>'."
      description = "Use kubectl to scale the specified deployment."
    },
    {
      name        = "Get Deployment Details"
      prompt      = "Ask the user for the deployment name. If a specific namespace is provided, filter deployments by that namespace using 'kubectl get deployments -n <namespace>'. Then, get details of the deployment using 'kubectl describe deployment/<deployment-name>'."
      description = "Use kubectl to get details of the specified deployment."
    }
  ]

  environment_variables = {
    DEBUG     = "1"
    LOG_LEVEL = "INFO"
  }
  
  is_debug_mode = false // boolean, optional
}

output "agent" {
  value = kubiya_agent.agent
}

Please note that before an agent is created we validate several inputs by making sure they exist and configured properly before handling your request. The validated inputs are: 1. runner 2. secrets 3. users 4. groups 5. integrations

Agent Parameters:

1. Mandatory Fields

  name         = "terraform-agent"                // The name of the agent
  runner      = "terraform"                      // The runner for the agent
  description  = "Terraform agent"                // The description of the agent
  instructions = "use terraform to run the agent" // The instructions for the agent

Optional Fields (omitting will retain the current values)

links    = []                       //Array

model = "azure/gpt-4"       // If not provided, Defaults to "azure/gpt-4"
//available options: "azure/gpt-4", "azure/gpt-4o", "azure/gpt-4-turbo-preview"
//"azure/gpt-4-32k, "azure/gpt-3.5-turbo

image = "ghcr.io/kubiyabot/kubiya-agent:stable" 
// Defaults to "ghcr.io/kubiyabot/kubiya-agent:stable"

// optional, the below inputs will be validated:
integrations = ["github","jira","aws"] //Array
users        = ["[e-mail]"] // Array
secrets      = ["terraform_secret","aws_secret"] // Array
groups       = ["Admin","terraformAdmins","terraformUsers"] // Array

starters = [
    {
      name = ""
      command      = ""
    }
]
//To remove, please use an empty block: starters = [].

 tasks = [
    {
      name        = ""
      prompt      = ""
      description = ""
    }
]
//To remove, please use an empty block: tasks = [].


environment_variables = {
    KEY = "VAL"
    KEY2 = "VAL2"
  }
//To remove, please use an empty block: environment_variables = {}.

is_debug_mode = true //Enable Agent Debug mode