ai.matey.types

ai.matey.types

TypeScript type definitions for the ai.matey ecosystem

Part of the ai.matey monorepo.

Installation

npm install ai.matey.types

Exports

Core IR Types

• IRChatRequest - Universal chat request format
• IRChatResponse - Universal chat response format
• IRMessage - Individual message in conversation
• IRStreamChunk - Streaming response chunk
• IRParameters - Generation parameters (temperature, maxTokens, etc.)
• IRMetadata - Request/response metadata and provenance
• IRTool - Tool/function definition
• IRUsage - Token usage statistics

Adapter Interfaces

• FrontendAdapter - Interface for frontend adapters
• BackendAdapter - Interface for backend adapters
• AdapterMetadata - Adapter capability metadata

Middleware

• Middleware - Middleware interface
• MiddlewareContext - Middleware execution context

Streaming

• StreamMode - Streaming mode ('delta' | 'accumulated')
• StreamingConfig - Streaming configuration options

Utilities

• MessageRole - Message role type
• MessageContent - Content block types
• FinishReason - Generation completion reasons
• IRCapabilities - Adapter capabilities
• IRWarning - Semantic drift warnings

Usage

import {
  IRChatRequest,
  IRChatResponse,
  IRMessage,
  IRStreamChunk,
  FrontendAdapter,
  BackendAdapter,
  Middleware
} from 'ai.matey.types';

// Create a chat request
const request: IRChatRequest = {
  messages: [
    { role: 'user', content: 'Hello!' }
  ],
  parameters: {
    model: 'gpt-4',
    temperature: 0.7
  },
  metadata: {
    requestId: 'req_123',
    timestamp: Date.now(),
    provenance: { frontend: 'openai' }
  }
};

Documentation

For comprehensive documentation of the IR format, see:

• IR Format Guide - Complete specification with examples
• API Reference - Full API documentation
• Type Definitions - Authoritative TypeScript source

License

MIT - see LICENSE for details.