@org.ai/types

Foundation type definitions for the AI primitives ecosystem.

Note for Consumers: This is an internal foundation package. For application development, use org.ai instead, which re-exports all types from this package plus additional organizational types (Role, Team, Goal, KPI, OKR).

Package Relationship

@org.ai/types (foundation - internal)
└── Provides base types used by all packages

org.ai (consumer-facing)
├── Re-exports ALL of @org.ai/types
└── Adds organizational types (Role, Team, Goal, KPI, OKR)

Consumers should:
  import { Thing, Agent, Role, Team } from 'org.ai'  // Correct

NOT:
  import { Thing, Agent } from '@org.ai/types'       // Internal use only

When to Use This Package

• Internal packages in the monorepo that need foundation types
• Building new packages that extend the type system

For application development, always use org.ai:

// Recommended for applications
import type { Thing, Agent, Human, Role, Team, Goal } from 'org.ai'

Foundation Types

This package provides the core type definitions:

Entity Types

• Thing, ThingDO - Base entity types with URL-based identity
• Things, ThingsDO, Collection - Collection types
• Noun, Verb, StandardVerbs - Schema types

Worker Types

• WorkerType - Base worker interface
• AgentType - AI agent with model, autonomy settings
• HumanType - Human worker with role, department, availability

Event Types (5W+H)

• Event - Full event with all dimensions
• EventWhat, EventWho, EventWhen, EventWhere, EventWhy, EventHow

Tool Types

• ToolType - Tool definition
• ToolParameterType, ToolInputType, ToolOutputType
• ExecutableToolType, ValidatableToolType
• ToolsType, ToolboxType

Business Framework Types

• LeanCanvasType - 9-box business model canvas
• StoryBrandType - 7-part narrative framework
• FounderType - Founding team member
• StartupType - Startup with lifecycle stages
• ICPType - Ideal Customer Profile

Utility Types

• AIFunctionType<TOutput, TInput, TConfig> - Generic AI function
• EventHandlerType<TOutput, TInput> - Workflow event handlers
• WorkflowContextType - Workflow execution context
• RelationshipOperatorType - Database relationship operators
• ParsedFieldType - Schema field definitions
• ListOptions, ListResult, PaginationInfo - Pagination types

Type Guards

• isWorker, isAgent, isHuman
• isTool, isToolParameter, isToolExecutionResult
• isStartup, isICP
• isLeanCanvas, isStoryBrand, isFounder

Factory Functions

• createAgent(opts) - Create an AI agent
• createHuman(opts) - Create a human worker

Installation (for internal packages)

pnpm add @org.ai/types

Usage (for internal packages)

import type {
  Thing,
  AgentType,
  HumanType,
  WorkflowContextType,
} from '@org.ai/types'

import { isAgent, createAgent } from '@org.ai/types'

Type Reference

| Type | Purpose | |------|---------| | AIFunctionType<TOutput, TInput, TConfig> | Generic AI function with output-first parameter order | | EventHandlerType<TOutput, TInput> | Workflow event handlers with typed context | | WorkflowContextType | The $ workflow proxy with send, try, do, on, every | | RelationshipOperatorType | Database relationship operators (->, ~>, <-, <~) | | ParsedFieldType | Schema field definitions with relationship metadata |

Why Output-First Generics?

Just like Promise<T> puts the important part first, AIFunctionType<TOutput> leads with what matters most:

// Natural reading order: "an AI function that returns a Summary"
type Summarizer = AIFunctionType<Summary>

// vs the awkward alternative
type Summarizer = AIFunctionType<Document, Summary>  // Wait, which is which?

License

MIT