Building AI Customer Service with N8N: Scraping Travel Site Data for Knowledge Bases

Overview

In the previous post the author introduced Retrieval‑Augmented Generation (RAG) and demonstrated the FastGPT UI for RAG. This fourth installment focuses on the first step of building an AI customer‑service system: preparing data and cleaning knowledge before importing it.

Goal

Use N8N, an open‑source workflow automation platform, to automatically collect scenic‑spot information from the Chinese travel website Ly.com , format the data, and store it in a FastGPT knowledge base, thereby enabling AI‑driven Q&A for travel attractions.

Workflow Diagram

Resulting Data Format

About N8N

N8N is an open‑source workflow automation platform with native AI support. It combines a visual drag‑and‑drop builder with code extensibility. As of 27 May 2025 it has 99 K stars on GitHub. Core capabilities include:

Visual builder + code extension : connect 400+ app nodes, support conditional branches and error retries.

Multi‑step AI agents : build complex AI workflows using LangChain.

Private deployment : one‑click Docker deployment, fully offline operation.

Code embedding : insert JavaScript or Python snippets at any step.

Debug tools : step‑by‑step retry, data simulation, real‑time logs.

Enterprise security : SSO, RBAC, audit logs, encrypted storage.

Template marketplace : 1 200+ ready‑made workflow templates.

Installation

docker volume create n8n_data</code><code>docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n

After the container starts, open http://localhost:5678 in a browser, register a user, and access the N8N dashboard.

Building the Workflow

The workflow consists of four nodes:

Webhook node : receives external requests, acting as the workflow’s entry point.

HTTP request node : fetches the scenic‑spot page from https://www.ly.com/scenery/BookSceneryTicket_32291.html.

AI node : parses the HTML and extracts structured information using a prompt (see below).

HTTP request node : sends the extracted data to FastGPT’s knowledge‑base API.

Each node is configured through the visual editor; screenshots in the original article illustrate the exact UI steps.

System Prompt for the AI Node

## Role
You are an expert in extracting and analyzing HTML page content. Extract the scenic‑spot title, level, description, address, and transportation info.
## Extraction Rules
1. Title is in the element with class "s_name".
2. Description is in class "s_des".
3. Address is in class "s_com s_address".
4. Additional info is in class "inf-f-con"; include images as Markdown if present.
5. Transportation details are in class "tra_line".
6. Return a Markdown‑formatted article without 
 characters.
## Output Format
### Scenic Spot Introduction
**Name**: 
**Level**: 
**Description & Images**
**Address**: 
**Transportation**:

FastGPT Integration

Steps to push data into FastGPT:

Create an API key in FastGPT.

Create a knowledge base and note its ID.

Build a JSON payload containing the Markdown result from the AI node.

Configure the second HTTP request node with the FastGPT endpoint, add the API key in the header, and set the request body to the JSON payload.

When the workflow is triggered (e.g., via Postman), the data flows through the nodes, and the final HTTP request writes the structured knowledge into FastGPT. The article shows screenshots of successful responses and the populated knowledge base.

Advantages of N8N

Open‑source & free : full data sovereignty, Docker‑based private deployment eliminates leakage risk.

Zero‑cost entry : self‑hosted without licensing fees.

400+ integrations : covers databases, cloud storage, hardware devices, enabling cross‑system automation.

Visual + code engine : suitable for both non‑technical users (drag‑and‑drop) and developers (custom JavaScript/Python).

Developer‑friendly : custom logic, micro‑service integration, reduces development cycles.

Drawback : learning curve requires understanding of APIs and workflow concepts.

Conclusion

The author demonstrates that combining N8N with FastGPT automates the entire pipeline—from web scraping to knowledge‑base ingestion—greatly improving efficiency and reducing manual errors in building an AI‑driven travel‑site customer service.