Model Entity

The Model entity defines AI/LLM service configurations for use in Prompt entities. Models store authentication credentials and endpoint URLs for calling language model services.

Overview

The platform supports two model systems:

1. Universal Model System (recommended) — A structured hierarchy of providers, definitions, credentials, and configurations supporting OpenAI, Anthropic, AWS Bedrock, Google, and xAI
2. Legacy Models — Direct URL + credential storage for backward compatibility

Universal Model System

The universal model system uses Model Configurations that combine a model definition with credentials and default parameters. See Settings - Universal Model System for details on creating and managing configurations.

Supported providers:

Provider	Models	Auth
OpenAI	GPT-4o, GPT-4o-mini, o1, o3-mini	API Key
Anthropic	Claude 3.5 Sonnet, Claude 3 Haiku, Claude 3 Opus	API Key
AWS Bedrock	Claude, Nova, Titan (via IAM)	IAM Role
Google	Gemini Pro, Gemini Flash	API Key
xAI	Grok-2, Grok-2-mini	API Key

Image recognition is supported for vision-capable models (GPT-4o, Claude 3.5 Sonnet, Gemini Pro Vision). The adapters handle provider-specific image formatting automatically.

Legacy Models

Legacy models store the URL and credentials directly. They support:

• Bearer token authentication (OpenAI, Anthropic direct)
• Client key/secret authentication (custom APIs)
• AWS Bedrock via special bedrock:// URL format (IAM auth)

Model Types

User Models

Organization-specific models created by users:

Field	Required	Description
organizationId	Yes	Owner organization UUID
name	Yes	Display name
modelUrl	Yes	API endpoint or bedrock://model-id
token	Conditional	Bearer token (OR clientKey/secretKey)
clientKey	Conditional	Client key (requires secretKey)
secretKey	Conditional	Secret key (requires clientKey)

System Models

Platform-wide models managed by admins:

Field	Required	Description
organizationId	No	NULL for system models
modelType	Yes	Set to "system"
isVisible	Yes	Show in org model selectors
name	Yes	Display name
modelUrl	Yes	API endpoint or bedrock://model-id

---

API Endpoints

Operation	Method	Endpoint	Permission
List Models	GET	/api/models	models:read
Create Model	POST	/api/models	models:create
Get Model	GET	/api/models/{id}	models:read
Update Model	PUT	/api/models/{id}	models:update
Delete Model	DELETE	/api/models/{id}	models:delete

---

Create User Model

POST /api/models
Content-Type: application/json

{
  "organizationId": "550e8400-e29b-41d4-a716-446655440000",
  "name": "OpenAI GPT-4o",
  "description": "Production OpenAI model",
  "modelUrl": "https://api.openai.com/v1/chat/completions",
  "token": "sk-proj-xxxxxxxxxxxxx"
}

Response:

{
  "id": "770e8400-e29b-41d4-a716-446655440001",
  "organizationId": "550e8400-e29b-41d4-a716-446655440000",
  "name": "OpenAI GPT-4o",
  "description": "Production OpenAI model",
  "modelUrl": "https://api.openai.com/v1/chat/completions",
  "modelType": "user",
  "isVisible": true,
  "hasToken": true,
  "hasClientKey": false,
  "createdAt": "2025-12-15T10:00:00.000Z",
  "updatedAt": "2025-12-15T10:00:00.000Z"
}

Security

Token, clientKey, and secretKey are never returned in API responses. Only hasToken and hasClientKey boolean flags are returned.

---

Create Bedrock Model

AWS Bedrock models use special bedrock:// URL format and IAM authentication:

POST /api/models
Content-Type: application/json

{
  "organizationId": "550e8400-e29b-41d4-a716-446655440000",
  "name": "Claude 3 Sonnet",
  "description": "AWS Bedrock Claude 3 Sonnet",
  "modelUrl": "bedrock://anthropic.claude-3-sonnet-20240229-v1:0",
  "token": "iam"
}

Bedrock URL Format:

bedrock://<model-id>

Supported Bedrock Models:

Provider	Model ID
Anthropic Claude 3 Sonnet	anthropic.claude-3-sonnet-20240229-v1:0
Anthropic Claude 3 Haiku	anthropic.claude-3-haiku-20240307-v1:0
Amazon Nova Pro	amazon.nova-pro-v1:0
Amazon Nova Lite	amazon.nova-lite-v1:0
Amazon Titan Text	amazon.titan-text-express-v1

IAM Authentication

For Bedrock models, the token field value is ignored. Authentication uses IAM roles configured on the ECS task. Set token to any non-empty value (e.g., "iam") to satisfy validation.

---

Create System Model

System models are available to all organizations:

POST /api/models
Content-Type: application/json

{
  "name": "Platform Claude 3",
  "description": "Shared Bedrock model for all orgs",
  "modelUrl": "bedrock://anthropic.claude-3-sonnet-20240229-v1:0",
  "token": "iam",
  "modelType": "system",
  "isVisible": true
}

---

List Models

Get Organization Models

GET /api/models?organizationId=550e8400-e29b-41d4-a716-446655440000

Get Organization + System Models

Use for model selection dropdowns:

GET /api/models?organizationId=550e8400-e29b-41d4-a716-446655440000&includeSystem=true

Get Only System Models

For admin management:

GET /api/models?modelType=system

---

Update Model

PUT /api/models/{id}
Content-Type: application/json

{
  "name": "Updated Model Name",
  "description": "Updated description"
}

To update credentials, include the new token or keys:

PUT /api/models/{id}
Content-Type: application/json

{
  "token": "new-api-key-here"
}

---

Using Models with Prompts

Attach Model to Prompt Entity

When creating a Prompt entity, reference a model by ID:

POST /api/workflow-entities
Content-Type: application/json

{
  "organizationId": "550e8400-e29b-41d4-a716-446655440000",
  "environmentId": "660e8400-e29b-41d4-a716-446655440001",
  "name": "Generate Tweet",
  "workflowEntityTypeId": "<prompt-type-uuid>",
  "modelId": "770e8400-e29b-41d4-a716-446655440001",
  "prompt": "Generate a tweet about: {{content}}",
  "tfCondition": "Single Path"
}

Model Execution at Runtime

When a Prompt entity with a model is executed:

1. Consumer retrieves model credentials from cache
2. Model URL determines authentication method:
   • bedrock:// → Uses IAM authentication
   • https:// + token → Bearer token auth
   • https:// + clientKey/secretKey → Custom header auth
3. Prompt is sent to model endpoint
4. Response stored via latestPromptResponse()

---

Best Practices

Security

1. Never share tokens — Each org should have its own API keys
2. Use Bedrock for production — IAM roles more secure than tokens
3. Rotate credentials — Update tokens periodically

Organization

1. Descriptive names — Include provider and capability: "OpenAI GPT-4o for Tweets"
2. Use system models — For shared platform capabilities
3. Hide deprecated — Set isVisible: false on old models

Cost Management

1. Choose appropriate model — Use cheaper models for simple tasks
2. Monitor usage — Track which models are called most
3. Set limits — Use environment variables to cap requests

---

Related Topics

• Prompt Entity — Uses models for AI execution
• Prompt Scripts — Runtime model execution functions
• AWS Bedrock Support — Bedrock-specific details