@aikit/core - AI Agent Orchestration Framework

A powerful, type-safe framework for building AI agents with tool calling capabilities, multi-step reasoning, and streaming support.

Features

• Type-Safe Tool Definitions: Use Zod schemas for robust parameter validation
• Multi-Step Execution: Automatic orchestration of LLM → Tool → LLM loops
• Streaming Support: Real-time execution updates and streaming responses
• Framework-Agnostic: Works with OpenAI, Anthropic, and custom LLM providers
• Execution Tracing: Detailed logging and debugging of agent execution
• Error Handling: Graceful error recovery with configurable retries
• Production-Ready: Comprehensive test coverage (80%+)

Installation

npm install @ainative-studio/aikit-core zod

Quick Start

Basic Agent

import { Agent, AgentExecutor } from '@ainative-studio/aikit-core';
import { z } from 'zod';

// Define your agent
const agent = new Agent({
  id: 'assistant',
  name: 'My Assistant',
  systemPrompt: 'You are a helpful AI assistant.',
  llm: {
    provider: 'openai',
    model: 'gpt-4',
    apiKey: process.env.OPENAI_API_KEY,
  },
  tools: [],
});

// Execute
const executor = new AgentExecutor(agent);
const result = await executor.execute('Hello! How are you?');

console.log(result.response);

See README.md for complete documentation with examples of:

• Agent with Tools
• Streaming Execution
• Error Handling
• Custom LLM Providers
• Multi-Agent Systems

API Reference

See full documentation in README.md

License

MIT