Generate Structured Metadata from Trustpilot Reviews

Name: Generate Structured Metadata from Trustpilot Reviews
Rating: 5 (5 reviews)
Author: Free N8N

Intermediate

13 nodes connected

detail.loadingPreview

Free N8N Temples

170 views

0 downloads

Data Processingautomationclusteringgoogle sheetslangchainmetadata generationqdrantscrapingtrustpilotvector database

Scrape Trustpilot reviews, process them, store them in a vector database, and generate structured metadata for analysis.

About This Workflow

This workflow automates the process of gathering customer feedback from Trustpilot. It scrapes reviews for a specified company, extracts key information, stores these reviews in a Qdrant vector database for similarity search, and then uses clustering to identify common themes and sentiments. Finally, it exports processed data to Google Sheets and can trigger further analysis workflows.

Key Features

Automated Scraping: Fetches Trustpilot reviews for a given company.
Data Extraction: Parses review data including author, rating, text, dates, and country.
Vector Storage: Stores reviews and their embeddings in Qdrant for efficient similarity search.
Text Splitting & Embedding: Prepares review text for embedding and processing.
Clustering Analysis: Groups similar reviews using vector embeddings to identify common feedback patterns.
Data Export: Outputs processed insights and raw reviews to Google Sheets.
Dynamic Date Filtering: Allows filtering reviews by a specific date range.
Workflow Triggering: Enables calling this workflow as a sub-workflow or a trigger.

How To Use

Trigger the workflow: This workflow can be triggered manually by clicking 'Test workflow' or via a workflow trigger.
Set Company ID: The 'Set Variables' node (ID: f0ea6b63-c96d-4b3f-8a21-d0f2dbb4efc3) defines the companyId for which reviews will be scraped. Ensure this is set correctly.
Initial Data Cleaning: The 'Clear Existing Reviews' node (ID: 1f60c3a5-a47a-4313-9b29-8ea652d573f7) ensures a clean slate in the Qdrant database for the specified company.
Scrape Reviews: The 'Get TrustPilot Page' node (ID: 139ccadd-9135-4681-b2eb-403b8d8bd710) scrapes reviews from Trustpilot. The maxRequests: 3 parameter limits scraping to the most recent 3 pages.
Extract Review Data: The 'Extract Reviews' node (ID: 9290e116-c001-49d5-ae4c-d91cd246f2c2) parses the HTML content to extract individual review details.
Structure Review Data: The 'Zip Entries' node (ID: 00de989c-d9e9-4b42-b5db-7097800a6017) consolidates extracted review fields into a structured array.
Store in Vector Database: The 'Qdrant Vector Store' node (ID: a4f82a1b-5a76-46b6-a7a3-84ab09b46699) inserts the structured review data and their generated embeddings into the 'trustpilot_reviews' collection in Qdrant.
Find Reviews for Analysis: The 'Find Reviews' node (ID: 85cb48b1-0ab9-4f88-88f3-82fcfb041ebe) retrieves reviews from Qdrant based on companyId and a specified date range (current month by default).
Cluster Reviews: The workflow then proceeds to cluster these reviews to identify common themes and points of interest. This involves fetching points from Qdrant (Get Payload of Points, Clusters To List, Only Clusters With 3+ points), likely using a code node (not explicitly shown but implied by Sticky Note4) to perform the clustering algorithm.
Prepare Output: The 'Prep Output For Export' node (ID: 69bbd197-c78f-4dae-9300-fe23d4d49855) formats the clustered insights and raw review data for export.
Export to Sheets: The 'Export To Sheets' node (ID: d77daa23-6acf-4daa-bf4c-33da4d05a54c) appends the prepared data to a Google Sheet.
Trigger Sub-workflow (Optional): The 'Trigger Insights' node (ID: 61c3117c-757c-45dd-b9d5-1122b793be30) can be used to trigger another workflow for further analysis using the generated insights.

Apps Used

automation

clustering

google sheets

langchain

metadata generation

qdrant

scraping

trustpilot

vector database

Workflow JSON

{
  "id": "fef39898-009f-4560-9956-dd098279b599",
  "name": "Generate Structured Metadata from Trustpilot Reviews",
  "nodes": 13,
  "category": "Data Processing",
  "status": "active",
  "version": "1.0.0"
}

Note: This is a sample preview. The full workflow JSON contains node configurations, credentials placeholders, and execution logic.