Skip to content
Pika Platform

Inference Profile Names

Complete reference for understanding how Pika names AWS Bedrock inference profiles for cost tracking and allocation.

Overview

Section titled “Overview”

Pika automatically creates named inference profiles by copying AWS's built-in inference profiles. These named profiles allow you to track costs for specific models in AWS Cost Explorer, enabling detailed cost allocation and analysis.

Naming Convention

Section titled “Naming Convention”

Inference profile names follow this pattern:

{stackName}-{profileName}

Components

Section titled “Components”

stackName

  • The kebab-case project name from pika-config.ts
  • Corresponds to pika.projNameKebabCase in your configuration
  • Example: pika-test, customer-portal-api, pika-chat-dev

profileName

  • Human-readable model identifier
  • Currently hardcoded by Pika (user configuration in future release)
  • Examples: claude-sonnet-4, claude-haiku-4-5, claude-sonnet-4-5

Example Names

Section titled “Example Names”

Given this configuration:

export const pikaConfig: PikaConfig = {
    pika: {
        projNameL: 'pika',
        projNameKebabCase: 'pika',
        // ... other config
    }
};

When deployed with STAGE=test, the inference profiles will be named:

  • pika-test-claude-sonnet-4
  • pika-test-claude-haiku-4-5
  • pika-test-claude-sonnet-4-5

Currently Created Profiles

Section titled “Currently Created Profiles”

Pika automatically creates inference profiles for these models:

ModelProfile NameBase Model ID
Claude 4 Sonnetclaude-sonnet-4us.anthropic.claude-sonnet-4-20250514-v1:0
Claude 4.5 Haikuclaude-haiku-4-5us.anthropic.claude-haiku-4-5-20251001-v1:0
Claude 4.5 Sonnetclaude-sonnet-4-5us.anthropic.claude-sonnet-4-5-20250929-v1:0

Resource Tags

Section titled “Resource Tags”

Each inference profile is tagged with:

  1. Stack tags from config - All tags from stackTags.common and stackTags.pikaServiceTags in pika-config.ts
  2. Component tag - Automatically added: component: {ModelName}InferenceProfile

Example Tags

Section titled “Example Tags”

Given this stack configuration:

export const pikaConfig: PikaConfig = {
    stackTags: {
        common: {
            'Environment': 'prod',
            'CostCenter': 'engineering',
            'Project': 'CustomerPortal'
        },
        pikaServiceTags: {
            'Team': 'platform'
        }
    }
};

The pika-prod-claude-sonnet-4-5 inference profile will be tagged with:

{
    "Environment": "prod",
    "CostCenter": "engineering",
    "Project": "CustomerPortal",
    "Team": "platform",
    "component": "Claude4_5SonnetInferenceProfile"
}

Complete Examples

Section titled “Complete Examples”

Single Environment

Section titled “Single Environment”
export const pikaConfig: PikaConfig = {
    pika: {
        projNameKebabCase: 'my-chatbot'
    },
    stackTags: {
        common: {
            'app': 'chatbot',
            'env': 'production'
        }
    }
};

Resulting inference profile names:

  • my-chatbot-production-claude-sonnet-4
  • my-chatbot-production-claude-haiku-4-5
  • my-chatbot-production-claude-sonnet-4-5

Multi-Environment

Section titled “Multi-Environment”
export const pikaConfig: PikaConfig = {
    pika: {
        projNameKebabCase: 'customer-api'
    },
    stackTags: {
        common: {
            'Environment': '{stage}',
            'Project': 'CustomerAPI',
            'CostCenter': 'platform'
        }
    }
};

When deployed to different stages:

Dev (STAGE=dev):

  • customer-api-dev-claude-sonnet-4
  • customer-api-dev-claude-haiku-4-5
  • customer-api-dev-claude-sonnet-4-5

Production (STAGE=prod):

  • customer-api-prod-claude-sonnet-4
  • customer-api-prod-claude-haiku-4-5
  • customer-api-prod-claude-sonnet-4-5

ARN Format

Section titled “ARN Format”

Inference profile ARNs follow this format:

arn:aws:bedrock:{region}:{accountId}:application-inference-profile/{profileName}

Example:

arn:aws:bedrock:us-east-1:123456789012:application-inference-profile/pika-test-claude-sonnet-4-5

Determining Your Profile Names

Section titled “Determining Your Profile Names”

To predict your inference profile names:

  1. Find your stack name:

    • Look at pika.projNameKebabCase in pika-config.ts
    • Append the stage with a hyphen (e.g., -test, -prod)
    • Example: pika-test, customer-api-prod

  2. Add the profile name:

    • Use the hardcoded profile names: claude-sonnet-4, claude-haiku-4-5, or claude-sonnet-4-5
    • Join with a hyphen: {stackName}-{profileName}

  3. Example calculation:

    pika.projNameKebabCase = "my-api"
    STAGE = "prod"
    profileName = "claude-sonnet-4-5"
    

    Result: "my-api-prod-claude-sonnet-4-5"

Use Cases

Section titled “Use Cases”

Cost Tracking in AWS Cost Explorer

Section titled “Cost Tracking in AWS Cost Explorer”

Find costs for specific models:

  1. Filter by inference profile name
  2. Group costs by environment, project, or team
  3. Compare costs across different models

See Track AI Model Costs for detailed instructions.

Resource Organization

Section titled “Resource Organization”

Use tags to:

  • Group all inference profiles for a project
  • Filter by environment or cost center
  • Identify resources by team or owner

Cost Allocation Reports

Section titled “Cost Allocation Reports”

Create custom cost allocation reports that:

  • Break down AI costs by model
  • Compare environments (dev vs prod)
  • Track costs over time per model

See Also

Section titled “See Also”