ObjectStackObjectStack

Model Registry

Model Registry protocol schemas

AI Model Registry Protocol

Centralized registry for managing AI models, prompt templates, and model versioning.

Enables AI-powered ObjectStack applications to discover and use LLMs consistently.

Source: packages/spec/src/ai/model-registry.zod.ts

TypeScript Usage

import { ModelCapability, ModelConfig, ModelLimits, ModelPricing, ModelProvider, ModelRegistryEntry, ModelSelectionCriteria, PromptVariable } from '@objectstack/spec/ai';
import type { ModelCapability, ModelConfig, ModelLimits, ModelPricing, ModelProvider, ModelRegistryEntry, ModelSelectionCriteria, PromptVariable } from '@objectstack/spec/ai';

// Validate data
const result = ModelCapability.parse(data);

ModelCapability

Properties

PropertyTypeRequiredDescription
textGenerationboolean✅Supports text generation
textEmbeddingboolean✅Supports text embedding
imageGenerationboolean✅Supports image generation
imageUnderstandingboolean✅Supports image understanding
functionCallingboolean✅Supports function calling
codeGenerationboolean✅Supports code generation
reasoningboolean✅Supports advanced reasoning

ModelConfig

Properties

PropertyTypeRequiredDescription
idstring✅Unique model identifier
namestring✅Model display name
versionstring✅Model version (e.g., "gpt-4-turbo-2024-04-09")
providerEnum<'openai' | 'azure_openai' | 'anthropic' | 'google' | 'cohere' | 'huggingface' | 'local' | 'custom'>✅
capabilitiesObject✅
limitsObject✅
pricingObjectoptional
endpointstringoptionalCustom API endpoint
apiKeystringoptionalAPI key (Warning: Prefer secretRef)
secretRefstringoptionalReference to stored secret (e.g. system:openai_api_key)
regionstringoptionalDeployment region (e.g., "us-east-1")
descriptionstringoptional
tagsstring[]optionalTags for categorization
deprecatedboolean✅
recommendedForstring[]optionalUse case recommendations

ModelLimits

Properties

PropertyTypeRequiredDescription
maxTokensinteger✅Maximum tokens per request
contextWindowinteger✅Context window size
maxOutputTokensintegeroptionalMaximum output tokens
rateLimitObjectoptional

ModelPricing

Properties

PropertyTypeRequiredDescription
currencystring✅
inputCostPer1kTokensnumberoptionalCost per 1K input tokens
outputCostPer1kTokensnumberoptionalCost per 1K output tokens
embeddingCostPer1kTokensnumberoptionalCost per 1K embedding tokens

ModelProvider

Allowed Values

  • openai
  • azure_openai
  • anthropic
  • google
  • cohere
  • huggingface
  • local
  • custom

ModelRegistryEntry

Properties

PropertyTypeRequiredDescription
modelObject✅
statusEnum<'active' | 'deprecated' | 'experimental' | 'disabled'>✅
priorityinteger✅Priority for model selection
fallbackModelsstring[]optionalFallback model IDs
healthCheckObjectoptional

ModelSelectionCriteria

Properties

PropertyTypeRequiredDescription
capabilitiesstring[]optionalRequired capabilities
maxCostPer1kTokensnumberoptionalMaximum acceptable cost
minContextWindownumberoptionalMinimum context window size
providerEnum<'openai' | 'azure_openai' | 'anthropic' | 'google' | 'cohere' | 'huggingface' | 'local' | 'custom'>optional
tagsstring[]optional
excludeDeprecatedboolean✅

PromptVariable

Properties

PropertyTypeRequiredDescription
namestring✅Variable name (e.g., "user_name", "context")
typeEnum<'string' | 'number' | 'boolean' | 'object' | 'array'>✅
requiredboolean✅
defaultValueanyoptional
descriptionstringoptional
validationObjectoptional

On this page

TypeScript UsageModelCapabilityPropertiesModelConfigPropertiesModelLimitsPropertiesModelPricingPropertiesModelProviderAllowed ValuesModelRegistryEntryPropertiesModelSelectionCriteriaPropertiesPromptVariableProperties