Skip to main content

​
Overview

ExuluApp is the central class for the Exulu Intelligence Management Platform (IMP). It orchestrates all components of your AI agent infrastructure, including agents, contexts, tools, evaluations, and background workers.

​
Key features

Agent management

Register and manage AI agents from multiple providers (OpenAI, Anthropic, Google, Cerebras)

Context system

Configure semantic search contexts with vector embeddings

Background workers

Process long-running tasks with BullMQ workers

Express integration

Built-in API server with GraphQL and REST endpoints

​
Installation

npm install @exulu/backend

​
Quick start

import { ExuluApp } from "@exulu/backend";

// Create and configure the app
const app = new ExuluApp();
await app.create({
  config: {
    telemetry: { enabled: false },
    workers: { enabled: true },
    MCP: { enabled: false }
  }
});

// Initialize Express server
const expressApp = await app.express.init();
expressApp.listen(3000, () => {
  console.log("Exulu IMP running on port 3000");
});

​
Architecture

The ExuluApp follows a factory pattern where you instantiate the class and then call the async create() method to initialize all components: 
const app = new ExuluApp();
const initializedApp = await app.create({
  contexts: { /* your contexts */ },
  agents: [ /* custom agents */ ],
  tools: [ /* custom tools */ ],
  evals: [ /* custom evaluations */ ],
  rerankers: [ /* custom rerankers */ ],
  config: { /* configuration */ }
});

​
Core concepts

AI agents are the execution units that process requests. Exulu IMP includes default agents for major providers:
  • Anthropic (Claude Opus 4, Sonnet 4, Sonnet 4.5)
  • OpenAI (GPT-5, GPT-4.1, GPT-4o series)
  • Google (Vertex Gemini 2.5 Flash, Pro, 3 Pro)
  • Cerebras (Llama 3.3, GPT-OSS)
You can add custom agents by passing them to the agents array.
Contexts are semantic search indices that allow agents to query relevant information. Each context:
  • Has an associated vector database table
  • Defines embeddings generation logic
  • Can include multiple data sources
  • Automatically exposes a tool for agents to use
Tools extend agent capabilities by providing functions they can call. Default tools include:
  • Todo management tools
  • Perplexity search tools
  • Context query tools (auto-generated from contexts)
BullMQ workers process background jobs:
  • Agent request execution
  • Embeddings generation
  • Evaluation runs
  • Scheduled context source updates

​
Next steps

Configuration

Learn about all configuration options

API reference

Explore methods and properties