MindLogic LogoMindLogic Docs
/
Dashboard

πŸ€– AI Agent Workflow: Build an Automated Bilingual Researcher

In today's information explosion era, manually consulting web pages and compiling summaries is an extremely time-consuming task. With MindLogic's Plugin System and Node Logic, you can easily orchestrate a fully automated AI Agent right on your canvas, just like playing with Lego blocks.

This tutorial will guide you from scratch to build a super assistant that, given a URL or topic, automatically completes a Content Scraping -> Core Summarization -> Bilingual Translation workflow.

Scenario Overview

In this workflow, we will achieve the following:

  1. Trigger Node: Receives the target URL or topic from the user.
  2. Scraper Node (Web Scraper): Automatically visits the URL and extracts all the main body text from the webpage.
  3. Analysis Node (LLM Summarizer): Uses an OpenAI-compatible plugin to summarize the lengthy web content into 3 core arguments.
  4. Translation Node (LLM Translator): Calls another LLM plugin to translate these 3 core arguments into professional-level Chinese/English.

Node Orchestration Steps

Step 1: Set the Initial Trigger Node

Create an entity node and name it [Research Target]. In the Inspector, add an Input parameter:

  • URL: https://en.wikipedia.org/wiki/Theory_of_constraints

Step 2: Configure the Web Scraper Plugin

Create a new node and name it [Data Extraction]. Draw a connection from [Research Target] to this node.

  1. In the right-hand "Plugin" panel, select the Web Scraper plugin.
  2. In the plugin configuration's URL field, enter {{ node.inputs['URL'] }} to dynamically receive the upstream URL.
  3. Upon execution, the plugin will automatically extract the web page's text and save it to the node's properties (usually node.outputs['text']).

Step 3: LLM Core Summarization

Create a new node and name it [Core Insights]. Draw a connection from [Data Extraction] to this node.

  1. Select the LLM (OpenAI Compatible) plugin (or DeepSeek) from the "Plugin" panel.
  2. In the System Prompt, enter: You are a professional analyst. Please read the following long text carefully and summarize the 3 most core arguments without any fluff.
  3. In the User Prompt, enter: {{ node.inputs['text'] }}
  4. When the computation flows through this node, the LLM condenses the massive web text into a concise summary, stored in node.outputs['response'].

Step 4: Multi-language Translation

Finally, create a node and name it [Translated Report]. Draw a connection from [Core Insights] to this node.

  1. Select an LLM plugin again.
  2. Set the System Prompt: You are a native business translator. Translate the text into highly professional language.
  3. Enter User Prompt: {{ node.inputs['response'] }}.
  4. Advanced Tip: In the custom "Script" section at the bottom of the node, write the following code to display the final result via a popup: 
    let report = node.outputs['response'];
node.title = "Report Ready!";
system.alert(report);

Results & Value

Through this extremely clear visual topology, we have built an AI Agent with a closed loop of Perceive (Scrape) -> Think (Summarize) -> Act (Translate). You can always swap out the LLM API, modify prompts, or even add a final node to POST the data to your Slack/Teams group via HTTP.

Efficiency Boost: Compresses reading and translating tasks that would normally take 30 minutes into just 10 seconds after clicking the run button.