Actions Reference

Complete list of available pipeline action types and their configuration

Overview

Actions are the building blocks of pipelines in BudAI Foundry. This reference provides details on all available action types and their configuration options.

Action Structure

Every action in a pipeline has the following structure:

{
    "id": "unique-step-id",
    "action": "action_type",
    "params": {
        # Action-specific parameters
    },
    "depends_on": ["previous-step-id"],  # Optional
    "condition": "expression",            # Optional
    "retry": {                           # Optional
        "max_attempts": 3,
        "backoff_multiplier": 2
    }
}

Available Action Types

Log Action

Log messages during pipeline execution.

Type: log

Parameters:

• message (string, required) - The message to log
• level (string, optional) - Log level: info, warning, error, debug. Default: info

Example:

{
    "id": "logger",
    "action": "log",
    "params": {
        "message": "Processing started",
        "level": "info"
    }
}

Transform Action

Transform data between pipeline steps.

Type: transform

Parameters:

• operation (string, required) - Transform operation: map, filter, reduce, normalize, uppercase, lowercase
• field (string, optional) - Field to transform
• function (string, optional) - Custom transformation function

Example:

{
    "id": "transformer",
    "action": "transform",
    "params": {
        "operation": "normalize",
        "field": "text"
    }
}

API Call Action

Make HTTP requests to external APIs.

Type: api_call

Parameters:

• url (string, required) - API endpoint URL
• method (string, required) - HTTP method: GET, POST, PUT, DELETE, PATCH
• headers (object, optional) - HTTP headers
• body (object, optional) - Request body for POST/PUT/PATCH
• timeout (integer, optional) - Request timeout in seconds. Default: 30

Example:

{
    "id": "fetch_data",
    "action": "api_call",
    "params": {
        "url": "https://api.example.com/data",
        "method": "POST",
        "headers": {
            "Content-Type": "application/json",
            "Authorization": "Bearer {{secrets.api_token}}"
        },
        "body": {
            "query": "value"
        }
    }
}

Condition Action

Conditional branching based on expressions.

Type: condition

Parameters:

• condition (string, required) - Boolean expression to evaluate
• on_true (string, optional) - Step ID to execute if true
• on_false (string, optional) - Step ID to execute if false

Example:

{
    "id": "check_status",
    "action": "condition",
    "params": {
        "condition": "steps.health_check.output.status == 'healthy'",
        "on_true": "deploy",
        "on_false": "notify_error"
    }
}

Loop Action

Iterate over collections of data.

Type: loop

Parameters:

• items (string, required) - Collection to iterate over (can reference previous step output)
• action (string, required) - Action to execute for each item
• max_iterations (integer, optional) - Maximum number of iterations. Default: 100
• parallel (boolean, optional) - Execute iterations in parallel. Default: false

Example:

{
    "id": "process_items",
    "action": "loop",
    "params": {
        "items": "steps.fetch_data.output.items",
        "action": "process_single_item",
        "parallel": true
    }
}

Model Add Action

Add a model to the model registry.

Type: model_add

Parameters:

• model_uri (string, required) - Model URI (e.g., meta-llama/Llama-3.2-1B-Instruct)
• model_source (string, required) - Model source: hugging_face, custom, s3, gcs
• model_name (string, optional) - Custom model name
• metadata (object, optional) - Additional model metadata

Example:

{
    "id": "add_model",
    "action": "model_add",
    "params": {
        "model_uri": "meta-llama/Llama-3.2-1B-Instruct",
        "model_source": "hugging_face",
        "model_name": "llama-3.2-1b"
    }
}

Deployment Create Action

Create a new model deployment.

Type: deployment_create

Parameters:

• model_id (string, required) - Model ID to deploy
• cluster_id (string, required) - Target cluster ID
• deployment_name (string, required) - Name for the deployment
• replicas (integer, optional) - Number of replicas. Default: 1
• gpu_count (integer, optional) - GPUs per replica. Default: 0
• memory_gb (integer, optional) - Memory in GB
• cpu_cores (integer, optional) - CPU cores

Example:

{
    "id": "deploy_model",
    "action": "deployment_create",
    "params": {
        "model_id": "{{steps.add_model.output.model_id}}",
        "cluster_id": "cluster_prod",
        "deployment_name": "llama-deployment",
        "replicas": 3,
        "gpu_count": 1
    }
}

Cluster Health Action

Check cluster health status.

Type: cluster_health

Parameters:

• cluster_id (string, required) - Cluster ID to check
• timeout (integer, optional) - Health check timeout in seconds. Default: 30

Example:

{
    "id": "health_check",
    "action": "cluster_health",
    "params": {
        "cluster_id": "cluster_prod",
        "timeout": 60
    }
}

Email Action

Send email notifications.

Type: email

Parameters:

• to (string or array, required) - Recipient email address(es)
• subject (string, required) - Email subject
• body (string, required) - Email body (supports templates)
• cc (string or array, optional) - CC recipients
• attachments (array, optional) - File attachments

Example:

{
    "id": "notify_success",
    "action": "email",
    "params": {
        "to": "[e-mail]",
        "subject": "Deployment Successful",
        "body": "Deployment {{steps.deploy.output.deployment_id}} completed successfully"
    }
}

Notification Action

Send notifications to various channels.

Type: notification

Parameters:

• message (string, required) - Notification message
• channel (string, required) - Notification channel: slack, teams, webhook
• webhook_url (string, optional) - Webhook URL for webhook channel
• severity (string, optional) - Severity level: info, warning, error. Default: info

Example:

{
    "id": "notify_slack",
    "action": "notification",
    "params": {
        "message": "Deployment failed: {{steps.deploy.error}}",
        "channel": "slack",
        "severity": "error"
    }
}

Database Action

Interact with databases.

Type: database

Parameters:

• connection (string, required) - Database connection string
• operation (string, required) - Operation: insert, update, delete, query
• table (string, required) - Target table name
• data (object, optional) - Data for insert/update operations
• query (string, optional) - SQL query for query operation

Example:

{
    "id": "save_results",
    "action": "database",
    "params": {
        "connection": "postgres://user:pass@host:5432/db",
        "operation": "insert",
        "table": "deployment_logs",
        "data": {
            "deployment_id": "{{steps.deploy.output.deployment_id}}",
            "status": "success",
            "timestamp": "{{now}}"
        }
    }
}

Wait Action

Pause execution for a specified duration.

Type: wait

Parameters:

• duration (integer, required) - Wait duration in seconds
• reason (string, optional) - Reason for waiting (for logging)

Example:

{
    "id": "wait_for_warmup",
    "action": "wait",
    "params": {
        "duration": 60,
        "reason": "Waiting for model warmup"
    }
}

Listing Available Actions

Using SDK

from budai import BudClient

client = BudClient(api_key="your-key")

actions = client.actions.list()

for action in actions:
    print(f"{action.type}: {action.description}")
    print(f"  Parameters: {action.parameters}")

Using CLI

bud action list

Get details for a specific action:

bud action get <action-type>

Template Variables

Actions support template variables for dynamic values:

Step Outputs

"{{steps.step_id.output.field_name}}"

Pipeline Parameters

"{{params.parameter_name}}"

Secrets

"{{secrets.secret_name}}"

Built-in Variables

• {{now}} - Current timestamp
• {{pipeline_id}} - Current pipeline ID
• {{execution_id}} - Current execution ID

Retry Configuration

All actions support retry configuration:

{
    "id": "action_with_retry",
    "action": "api_call",
    "params": {...},
    "retry": {
        "max_attempts": 3,
        "backoff_multiplier": 2,
        "retry_on": ["timeout", "connection_error"]
    }
}

Conditional Execution

Execute actions conditionally:

{
    "id": "conditional_action",
    "action": "notification",
    "params": {...},
    "condition": "steps.deploy.status == 'failed'"
}

Next Steps

Pipelines & Executions

Learn how to use actions in pipelines

Pipeline DSL

Build pipelines with Python DSL