Introduction

• playground  You can see the demo code on the playground.

This tutorial guides you through setting up a Retrieval-Augmented Generation (RAG) Agent using Agentica. With the code below, you can create an agent that combines OpenAI’s GPT model with a vector store for document retrieval and summarization.

The Complete RAG Agent

First, install the required package:

npm install @agentica/openai-vector-store

Then, add the following code to your project:

import { Agentica } from "@agentica/core";
import { AgenticaOpenAIVectorStoreSelector } from "@agentica/openai-vector-store";
import dotenv from "dotenv";
import OpenAI from "openai";
import typia from "typia";
 
dotenv.config();
 
const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY });
 
const selector = new AgenticaOpenAIVectorStoreSelector({
  provider: {
    api: openai,
    assistant: { name: "agentica", model: "gpt-4o-mini" },
    vectorStore: { name: "vectorStore" },
  },
});
 
export const RagAgent = new Agentica({
  model: "chatgpt",
  vendor: {
    api: openai,
    model: "gpt-4o-mini",
  },
  controllers: [
    {
      protocol: "class",
      name: "vectorStore",
      application: typia.llm.application<
        AgenticaOpenAIVectorStoreSelector,
        "chatgpt"
      >(),
      execute: selector,
    },
  ],
});

How It Works

With this configuration, the RAG Agent will:

• Utilize a Vector Store Selector: The AgenticaOpenAIVectorStoreSelector connects OpenAI’s API with a vector store, enabling document retrieval based on natural language queries.
• Process Natural Language Queries: The agent converts user queries into vector-based searches, retrieving relevant documents.
• Generate Intelligent Responses: It then uses the GPT model to summarize or respond to queries based on the retrieved documents.
• Ensure Type Safety: By using typia, the code maintains strong type safety.
• Secure API Access: Environment variables (managed by dotenv) secure your API keys.

For more detailed information about AgenticaOpenAIVectorStoreSelector, please refer to the Vector Store repository .

Use Case Example

Imagine asking the agent:

“Summarize recent advancements in AI research.”

The agent will process your query by searching the vector store for relevant documents, and then use OpenAI’s GPT model to generate a concise summary based on the retrieved information.

Environment Setup

Create a .env file in your project root with the following content:

OPENAI_API_KEY=your-openai-api-key

Final Notes & Tips

• Testing and Debugging:

  • During development, log responses and verify that your queries return the expected results.
  • Consider adding error handling to manage potential issues like API timeouts or vector store response delays.

• Security and Environment Variables:

  • Ensure that your API keys in the .env file are not committed to version control.
  • For production environments, use a secure secret management system.

• Vector Store Functionality:

  • The vector store is key to mapping natural language queries to relevant documents. Understanding its role can help you optimize search results.

• Additional Use Cases:

  • Beyond summarization, the agent can support various document retrieval tasks such as Q&A, content aggregation, or even generating insights based on a corpus of documents.

This code block sets up a fully functional RAG Agent using Agentica. Enjoy harnessing the power of natural language interactions combined with vector-based document retrieval to create dynamic, intelligent responses!