Agentic

A TypeScript library for running AI models across multiple providers with a unified interface.

Overview

Agentic provides a simple, consistent API for interacting with various AI providers including Cloudflare Workers AI and Groq. It handles provider-specific execution logic, retry logic with exponential backoff, JSON mode, tool calling, and response formatting.

Installation

npm install @em3odme/agentic
pnpm install @em3odme/agentic
yarn add @em3odme/agentic

Requirements: Node.js >= 18.0.0

Quick Start

import { ModelRunner, groqAIModel } from '@em3odme/agentic';

const runner = new ModelRunner({
  GROQ_API_KEY: process.env.GROQ_API_KEY!,
});

const result = await runner.run({
  messages: [{ role: 'user', content: 'Hello, how are you?' }],
  model: groqAIModel('llama-3.3-70b-versatile'),
});

console.log(result.content);

Supported Providers

Provider	JSON Mode	Tools	Streaming	Setup
Cloudflare Workers AI	✓	✓	✗	AI binding
Groq	✓	✓	✓	GROQ_API_KEY

Cloudflare Workers AI

Run open-source models (Llama, DeepSeek, Mistral, etc.) on Cloudflare's global edge network.

import { ModelRunner, cloudflareAIModel } from '@em3odme/agentic';

const runner = new ModelRunner({ AI: env.AI });

const result = await runner.run({
  messages: [{ role: 'user', content: 'Hello!' }],
  model: cloudflareAIModel('@cf/meta/llama-3.3-70b-instruct-fp8-fast'),
});

Features:

• Serverless GPU infrastructure
• JSON mode for structured responses
• Function calling capabilities
• Embedding models support
• Automatic retries with exponential backoff

Groq

High-performance LLM inference with ultra-low latency.

import { ModelRunner, groqAIModel } from '@em3odme/agentic';

const runner = new ModelRunner({
  GROQ_API_KEY: process.env.GROQ_API_KEY!,
});

const result = await runner.run({
  messages: [{ role: 'user', content: 'Hello!' }],
  model: groqAIModel('llama-3.3-70b-versatile'),
});

Features:

• Real-time streaming responses
• JSON mode for structured outputs
• Tool/function calling
• Speech recognition (Whisper models)
• Ultra-fast inference

Usage Examples

JSON Mode

const result = await runner.run({
  messages: [{ role: 'user', content: 'List top 3 programming languages' }],
  model: groqAIModel('llama-3.3-70b-versatile'),
  jsonMode: true,
});

const languages = JSON.parse(result.content);

Tool Calling

const result = await runner.run({
  messages: [{ role: 'user', content: 'What is the weather in New York?' }],
  model: groqAIModel('llama-3.3-70b-versatile'),
  options: { tools: true },
});

console.log(result.tool_calls);

Runtime Configuration

ModelRunner.updateRuntimeConfig({
  timeout: 60000,
  retries: {
    maxAttempts: 5,
    baseDelay: 1000,
    maxDelay: 10000,
  },
});

Documentation

For detailed documentation, see:

• ModelRunner Documentation - Core API reference
• Cloudflare Provider - Cloudflare Workers AI setup and usage
• Groq Provider - Groq API setup and usage

Package Info

• Repository: https://github.com/Em3ODMe/agentic.git
• Issues: https://github.com/Em3ODMe/agentic/issues

Keywords

agent, llm, ai, typescript, nodejs