Skip to Content
Getting StartedQuick Start

Quick Start

Get a working agent running in under 5 minutes.

Step 1 — Define Your LLM Provider

LLM Layer Engine does not ship a built-in provider. You bring your own by implementing the LLMProvider interface. Here is a LangChain example using Claude:

import { ChatAnthropic } from "@langchain/anthropic";
import type { LLMProvider, LLMResponse, AgentConfig, Message } from "llm-layer-engine";
 
const model = new ChatAnthropic({ model: "claude-3-5-sonnet-20241022" });
 
const myProvider: LLMProvider = {
  name: "anthropic",
  async run({ messages, config }): Promise<LLMResponse> {
    const response = await model.invoke(
      messages.map((m) => ({ role: m.role, content: m.content }))
    );
 
    return {
      content: response.content as string,
    };
  },
};

Step 2 — Run a Simple Agent

import { createAgent } from "llm-layer-engine";
 
const agent = createAgent({
  provider: myProvider,
  model: "claude-3-5-sonnet-20241022",
  temperature: 0.5,
});
 
const { result, stats } = await agent.run([
  { role: "user", content: "What is the capital of France?" },
]);
 
console.log(result);  // Paris
console.log(stats);   // { durationMs, tokenUsage, ... }

Step 3 — Add a Tool

import { registerTool, reActLoop } from "llm-layer-engine";
 
registerTool({
  name: "get_weather",
  execute: async ({ city }) => {
    return { weather: "Sunny", city };
  },
});
 
const loop = reActLoop({
  config: { provider: myProvider, model: "claude-3-5-sonnet-20241022" },
  provider: myProvider,
  messages: [{ role: "user", content: "What is the weather in Karachi?" }],
  tools: [{ name: "get_weather", execute: async (i) => ({ weather: "Sunny", city: i.city }) }],
  maxSteps: 5,
});
 
const { result, stats } = await loop.run();
console.log(result);

What Happens Under the Hood (User Perspective)

  1. createAgent sends your messages to the provider and returns the result + stats
  2. reActLoop runs multiple steps — calling tools when needed — until it has a final answer
  3. registerTool adds your tool to the global registry so the agent can find it

That is everything you need to get started.

Next Step

Last updated on
InstallationcreateAgent