AI Orchestrator Core

AI Orchestrator Core is a TypeScript library designed to provide the essential building blocks for orchestrating AI agents, assistants, and large language model (LLM) workflows. It is built with modularity and extensibility in mind, following Clean Architecture principles, and is intended for use in advanced AI orchestration scenarios within Node.js/NestJS projects.

Table of Contents

• Overview
• Features
• Installation
• Getting Started
• Architecture
• Core Modules & Concepts
• Usage Examples
• Extending the Library
• License

Overview

AI Orchestrator Core provides the foundation for building, configuring, and running AI agents and assistants. It abstracts the orchestration of LLMs, agent logic, prompt management, and tool integration, making it easy to build scalable and maintainable AI-driven applications.

Features

• Modular orchestration of LLMs and agents
• Extensible base classes for agents, assistants, and orchestrators
• Generic and file-based agent/assistant implementations
• DTOs and controllers for REST API integration
• Clean separation of domain, application, infrastructure, and presentation layers

Installation

Install via npm:

npm install @sftech/ai-orchestrator-core

Or with yarn:

yarn add @sftech/ai-orchestrator-core

Getting Started

Import the module into your NestJS application:

import { AiOrchestratorCoreModule } from '@sftech/ai-orchestrator-core';

@Module({
  imports: [
    AiOrchestratorCoreModule.register({
      // Provide orchestrator options here (see below)
    }),
  ],
})
export class AppModule {}

Orchestrator Options

The register method expects an options object implementing IOrchestratorOptions. This allows you to configure the orchestrator's behavior, such as agent/assistant types, tool integrations, and more. See src/domain/orchestrator-options.interface.ts for all available options.

Example:

const orchestratorOptions: IOrchestratorOptions = {
  // Define your orchestrator configuration here
};

AiOrchestratorCoreModule.register(orchestratorOptions);

Architecture

The library is structured according to Clean Architecture principles:

• Domain: Core models, interfaces, and enums
• Application: Base and generic orchestrators, agents, assistants, and file-based logic
• Infrastructure: Mappers, seeders, and option handling
• Presentation: REST controllers and DTOs for API integration

Core Modules & Concepts

• Base Classes: Extend BaseOrchestrator, BaseAgent, BaseAssistant, or BaseLlmAssistant to implement custom logic
• Generic Implementations: Use or extend GenericLlmOrchestrator, GenericLlmAgent, GenericLlmAssistant for LLM-based workflows
• File Reader Agents/Assistants: Integrate file-based input/output via FileReaderAgent and FileReaderAssistant
• Controllers: Use GenericOrchestratorController and ToolsController for REST API endpoints
• DTOs: Data transfer objects for configuration and runtime (see src/presentation/dtos/)

Usage Examples

Registering the Module

import { AiOrchestratorCoreModule } from '@sftech/ai-orchestrator-core';
import { IOrchestratorOptions } from '@sftech/ai-orchestrator-core';

const options: IOrchestratorOptions = {
  // ...your orchestrator configuration...
};

@Module({
  imports: [AiOrchestratorCoreModule.register(options)],
})
export class AppModule {}

Using the REST API

The library provides controllers for orchestrator and tool operations. Once the module is registered, endpoints such as /orchestrator/generic and /tools are available for integration.

Extending Agents or Assistants

You can extend the base or generic classes to implement your own agent or assistant logic:

import { BaseAgent } from '@sftech/ai-orchestrator-core';

export class MyCustomAgent extends BaseAgent {
  // Implement your custom logic here
}

Extending the Library

• Add new agent or assistant types by extending the base classes
• Implement custom orchestrators for advanced workflows
• Add new controllers or DTOs for additional API endpoints
• Use the provided interfaces and enums for strong typing and consistency

License

© 2025 SFTech. All rights reserved.