How to create a Youtube video summariser agent using EnvoyJS

In this tutorial, I will explain how you can build a Youtube video summariser AI agent using EnvoyJS, a minimal agentic framework in Javascript.

The idea

We will create a youtube video summariser agent and instruct them about their job. We will supply YoutubeTranscript tool to the agent so it can fetch the transcript of a given video.

We can either use the gpt-4o model, or deep-seek for this.

Install EnvoyJS library

npm i @envoyjs/core

Import EnvoyJS library, and Youtube Transcript tool

import { Agent, youtubeTranscriptTool } from "@envoyjs/core";

Create the agent

Let's create our agent using the Agent class.

const agent = new Agent({
    name: "Content Summariser Agent",
    bio: "You are an expert in reading YouTube video transcripts and summarizing what the video is about.",
    steps: [
        "Read the transcript",
        "Understand the entire context",
        "Find relevant portions",
        "Summarize as text",
        "Save it as a file"
    ],
    modelConfig: {
        provider: "OPEN_AI",
        apiKey: "", // Don't forget to add your OpenAI API key
        model: "gpt-4o"
    },
    tools: [youtubeTranscriptTool, fileWriterTool] // Add tools here
});

In this class, we're giving instructions to the agent on how to do the task. We have given a bio, and then steps instructions on how to perform!

Running the agent!

Let's prompt the agent to do the work using the print_response() function.

agent.printResponse("Summarise this youtube video - https://www.youtube.com/watch?v=QtYEPYntfL4&t=27s");

Now sitback and see how the agent analyses the task, take decisions to use the tool to fetch the transcript and summarise the video for you!