Creating and Managing Teammates with Kubiya CLI

This guide provides detailed instructions for creating, configuring, and interacting with AI teammates using the Kubiya CLI.

Understanding Teammates

Teammates are AI assistants that help you automate tasks, answer questions, and provide guidance in your specific domain. Each teammate can be customized with different capabilities, knowledge, and access to specific tools.

Listing Teammates

To see all available teammates in your organization:

kubiya teammate list

This command displays a list of all teammates with their IDs, names, and descriptions.

For more detailed output in JSON format:

kubiya teammate list --output json

To view teammate capabilities and additional details:

kubiya teammate list --verbose

Viewing Teammate Details

To get detailed information about a specific teammate:

kubiya teammate describe TEAMMATE_ID

This shows the teammate's configuration, capabilities, and associated resources.

For JSON output:

kubiya teammate describe TEAMMATE_ID --output json

Creating a New Teammate

You can create teammates with specific capabilities tailored to your needs:

Basic Teammate Creation

kubiya teammate create --name "Infrastructure-Expert" \
                      --description "Specialist in AWS and Kubernetes infrastructure" \
                      --capabilities "aws,kubernetes,terraform" \
                      --knowledge-items "item1,item2" \
                      --sources "source1,source2"

Advanced Teammate Creation

kubiya teammate create --name "DevOps-Expert" \
                      --description "DevOps automation specialist" \
                      --capabilities "aws,kubernetes,terraform,jenkins" \
                      --knowledge-items "devops-best-practices,cicd-patterns" \
                      --sources "devops-tools,infra-tools" \
                      --llm-model "gpt-4" \
                      --instructions "You are a DevOps expert who helps with infrastructure automation." \
                      --env-vars "AWS_REGION=us-west-2,DEBUG=true" \
                      --integrations "aws,kubernetes,github" \
                      --secrets "aws-credentials,github-token"

Creating from JSON/YAML

You can also create teammates from a file:

kubiya teammate create --from-file teammate-definition.json

Example teammate-definition.json:

{
  "name": "Data-Scientist",
  "description": "Machine learning and data analysis expert",
  "capabilities": ["python", "ml", "pandas", "tensorflow"],
  "sources": ["data-science-tools"],
  "instructions": "You are a data science expert who helps with ML models and data analysis.",
  "llm_model": "gpt-4",
  "env_vars": {
    "PYTHONPATH": "/opt/ml/lib",
    "CUDA_VISIBLE_DEVICES": "0"
  },
  "integrations": ["aws-sagemaker", "databricks"]
}

Using stdin:

cat teammate-definition.json | kubiya teammate create --from-stdin

Required Parameters:

• --name: A unique name for your teammate

• --description: A description of the teammate's purpose and expertise

Optional Parameters:

• --capabilities: Comma-separated list of capabilities

• --knowledge-items: Knowledge items to associate with the teammate

• --sources: Sources containing tools the teammate can access

• --llm-model: Language model to use (e.g., "gpt-4", "claude-3")

• --instructions: Custom instructions for the teammate

• --avatar: URL or path to an avatar image

• --env-vars: Environment variables for the teammate (KEY=VALUE format)

• --integrations: Integrations to enable (aws, github, slack, etc.)

• --secrets: Secrets to make available to the teammate

Editing a Teammate

You can edit a teammate in your default text editor, which allows for more comprehensive changes:

kubiya teammate edit TEAMMATE_ID

This opens your default editor ($EDITOR) with the teammate's configuration in JSON format, allowing you to make changes and save them.

Updating a Teammate

To modify an existing teammate's configuration through command line:

kubiya teammate update TEAMMATE_ID \
                      --description "Updated description" \
                      --capabilities "aws,kubernetes,terraform,docker" \
                      --knowledge-items "item1,item2,item3"

Adding and Removing Resources

Add or remove sources:

kubiya teammate update TEAMMATE_ID --add-sources "new-source1,new-source2"
kubiya teammate update TEAMMATE_ID --remove-sources "old-source"

Add or remove environment variables:

kubiya teammate update TEAMMATE_ID --add-env-vars "KEY1=VALUE1,KEY2=VALUE2"
kubiya teammate update TEAMMATE_ID --remove-env-vars "OLD_KEY"

Add or remove secrets:

kubiya teammate update TEAMMATE_ID --add-secrets "new-secret1,new-secret2"
kubiya teammate update TEAMMATE_ID --remove-secrets "old-secret"

Add or remove integrations:

kubiya teammate update TEAMMATE_ID --add-integrations "github,jira"
kubiya teammate update TEAMMATE_ID --remove-integrations "slack"

Changing the LLM Model

kubiya teammate update TEAMMATE_ID --llm-model "gpt-4"

Updating Instructions

kubiya teammate update TEAMMATE_ID --instructions "You are a specialized teammate focused on cloud infrastructure."

Deleting a Teammate

To remove a teammate from your organization:

kubiya teammate delete TEAMMATE_ID

You'll be prompted to confirm the deletion.

With force option to skip confirmation:

kubiya teammate delete TEAMMATE_ID --force

Chatting with Teammates

Once you've created teammates, you can interact with them in various ways:

Interactive Chat

Start an interactive chat session with a specific teammate:

kubiya chat --teammate "Infrastructure-Expert" --interactive

This opens a chat session where you can have a conversation with the teammate.

One-off Questions

Ask a specific question without starting an interactive session:

kubiya chat --teammate "Infrastructure-Expert" "How do I set up an AWS S3 bucket?"

Providing Context

You can provide context to help the teammate understand your question better:

kubiya chat --teammate "Infrastructure-Expert" "What's wrong with this Terraform code?" --files "main.tf,variables.tf"

Or use stdin to pipe data:

cat error.log | kubiya chat --teammate "Infrastructure-Expert" "What's causing these errors?"

Managing Chat Sessions

Continuing Previous Sessions

To continue a previous conversation with a teammate:

# List active sessions
kubiya chat --list-sessions

# Continue a specific session
kubiya chat --interactive --session SESSION_ID

Clearing Session Context

To start fresh while keeping the same session:

kubiya chat --interactive --clear-session

Language Models and Capabilities

Kubiya supports various language models for teammates:

Available Models

# Set specific model when creating
kubiya teammate create --name "Advanced-Assistant" --llm-model "gpt-4"

# Common models include:
# - gpt-4
# - gpt-3.5-turbo
# - claude-3
# - claude-2
# - llama-2

Model-Specific Instructions

Different models might require different instruction formats. The CLI handles these differences automatically:

kubiya teammate create --name "GPT-Teammate" --llm-model "gpt-4" --instructions "You are a helpful assistant."
kubiya teammate create --name "Claude-Teammate" --llm-model "claude-3" --instructions "You are a helpful assistant."

Environment Variables and Integrations

Setting Environment Variables

Environment variables provide configuration to your teammates:

kubiya teammate create --name "Cloud-Expert" \
                      --env-vars "AWS_REGION=us-west-2,LOG_LEVEL=debug"

Managing Integrations

Enable integrations with external systems:

kubiya teammate create --name "DevTools-Expert" \
                      --integrations "github,jira,slack"

Available integrations include:

• aws: Amazon Web Services

• kubernetes or k8s: Kubernetes clusters

• github: GitHub repositories

• gitlab: GitLab repositories

• jira: Jira issue tracking

• slack: Slack workspace

• teams: Microsoft Teams

• datadog: Datadog monitoring

• prometheus: Prometheus metrics

• jenkins: Jenkins CI/CD

• terraform: Terraform infrastructure

Creating Specialized Teammates

Here are examples of creating teammates for specific roles:

DevOps Teammate

kubiya teammate create --name "DevOps-Expert" \
                      --description "Specializes in CI/CD, infrastructure as code, and container orchestration" \
                      --capabilities "aws,kubernetes,terraform,jenkins,gitlab,docker" \
                      --sources "devops-tools,kubernetes-tools" \
                      --env-vars "DEFAULT_REGION=us-west-2,KUBE_VERSION=1.25" \
                      --integrations "aws,kubernetes,github,jenkins" \
                      --llm-model "gpt-4" \
                      --instructions "You are a DevOps expert who excels at automating infrastructure tasks, CI/CD pipelines, and Kubernetes management."

Security Teammate

kubiya teammate create --name "Security-Expert" \
                      --description "Focused on security best practices, vulnerability assessment, and compliance" \
                      --capabilities "security,compliance,aws-security,kubernetes-security" \
                      --knowledge-items "security-best-practices,compliance-frameworks" \
                      --integrations "aws,vault" \
                      --llm-model "claude-3" \
                      --instructions "You are a cybersecurity expert focused on cloud security, compliance, and vulnerability management."

Database Teammate

kubiya teammate create --name "Database-Expert" \
                      --description "Specializes in database management, optimization, and troubleshooting" \
                      --capabilities "mysql,postgresql,mongodb,redis,dynamodb" \
                      --sources "database-tools" \
                      --env-vars "DEFAULT_ENGINE=postgresql,DEBUG_SQL=true" \
                      --llm-model "gpt-4" \
                      --instructions "You are a database expert who helps with query optimization, schema design, and database administration."

Best Practices

1. Create Specialized Teammates: Rather than having one general-purpose teammate, create multiple specialized teammates for different domains.

2. Provide Detailed Descriptions: Include clear descriptions when creating teammates to help users understand their capabilities.

3. Associate Relevant Sources: Connect teammates with the appropriate tool sources to enhance their capabilities.

4. Utilize Knowledge Items: Add relevant knowledge items to teammates to improve their domain expertise.

5. Use Environment Variables: Use environment variables to configure teammate behavior and default settings.

6. Choose the Right Model: Select appropriate LLM models based on the teammate's purpose and complexity.

7. Provide Custom Instructions: Give detailed instructions to shape the teammate's behavior and response style.

8. Enable Relevant Integrations: Only enable the integrations that the teammate needs for its role.

9. Start with Templates: Use pre-built templates as a starting point for common teammate roles.

kubiya teammate create --from-template "devops"

End-to-End Example: Creating and Using a Cloud Infrastructure Teammate

Here's a complete workflow for creating and interacting with a specialized cloud infrastructure teammate:

Step 1: Create a New Teammate

kubiya teammate create --name "Cloud-Expert" \
                      --description "Specialist in AWS, Azure, and GCP infrastructure" \
                      --capabilities "aws,azure,gcp,terraform,pulumi" \
                      --sources "cloud-tools" \
                      --llm-model "gpt-4" \
                      --env-vars "DEFAULT_CLOUD=aws,DEFAULT_REGION=us-west-2" \
                      --integrations "aws,azure,gcp" \
                      --instructions "You are a cloud infrastructure expert specialized in multi-cloud environments. You help with designing, deploying, and troubleshooting cloud resources across AWS, Azure, and GCP."

Step 2: Get the Teammate ID

TEAMMATE_ID=$(kubiya teammate list --output json | jq -r '.[] | select(.name=="Cloud-Expert") | .id')

Step 3: Add Knowledge Items to the Teammate

kubiya knowledge create --title "AWS Best Practices" --content "aws-practices.md"
kubiya knowledge create --title "Multi-Cloud Strategy" --content "multi-cloud.md"
kubiya teammate update $TEAMMATE_ID --add-knowledge-items "AWS Best Practices,Multi-Cloud Strategy"

Step 4: Add Necessary Secrets

kubiya secret create --name "aws-credentials" --from-file ~/.aws/credentials --description "AWS credentials file"
kubiya teammate update $TEAMMATE_ID --add-secrets "aws-credentials"

Step 5: Review Teammate Configuration

kubiya teammate describe $TEAMMATE_ID

Step 6: Start an Interactive Chat

kubiya chat --teammate "Cloud-Expert" --interactive

Step 7: Ask Specific Questions with Context

kubiya chat --teammate "Cloud-Expert" "How can I optimize this infrastructure?" --files "terraform/*.tf"

Step 8: Update the Teammate as Needed

If you need to adjust the teammate's capabilities or configuration:

kubiya teammate edit $TEAMMATE_ID

Or via command line:

kubiya teammate update $TEAMMATE_ID --add-capabilities "terraform-cloud,cloudformation"

For more information about specific teammate capabilities and configuration options, see the official documentation.