Scrape Trustpilot Reviews, Extract Structured Metadata with DeepSeek, Analyze Sentiment with OpenAI, and Store in Google Sheets

• Automated Review Scraping: Fetches reviews from Trustpilot for a given company and number of pages.
• Structured Metadata Extraction: Utilizes DeepSeek (via Langchain) to extract specific fields like author, rating, date, title, text, number of reviews, and country.
• Sentiment Analysis: Employs OpenAI (via Langchain) to determine the sentiment (Positive, Neutral, Negative) of each review.
• Duplicate Checking: Integrates with Google Sheets to check if a review has already been processed.
• Data Storage: Appends or updates processed reviews, along with their extracted metadata and sentiment, into a Google Sheet.
• Configurable Parameters: Allows setting the company ID and maximum number of pages to scrape.

1. Configure Company and Pages: Update the Set Parameters node (company_id and max_page) with the target company's Trustpilot ID and the desired number of pages to scrape.
2. Set Up Google Sheets Credentials: Ensure your Google Sheets OAuth2 credentials are correctly configured in n8n and are linked to the Get rows, Get Google Sheets, and Update sheet nodes.
3. Specify Google Sheet: In the Get rows, Get Google Sheets, and Update sheet nodes, ensure the documentId and sheetName correctly point to your target Google Sheet.
4. Configure LLM Credentials: Ensure your DeepSeek and OpenAI API keys are set up correctly in n8n credentials and linked to the respective DeepSeek Chat Model and OpenAI Chat Model nodes.
5. Run the Workflow: Execute the workflow manually or set up a schedule.

Node Explanations:

• When clicking ‘Test workflow’: Manual trigger to start the workflow.
• Set Parameters: Defines the company to scrape and the maximum number of pages.
• Get reviews (HTTP Request): Fetches the HTML content of the Trustpilot review page.
• Extract (HTML): Extracts review links from the fetched HTML.
• Split Out: Splits the array of review links into individual items.
• Limit1: Limits the number of reviews processed to a maximum of 3 (can be adjusted).
• Get Google Sheets: Checks if a review already exists in the Google Sheet based on its ID.
• If: Conditional node to proceed only if the review is not found in Google Sheets.
• Get Single review (HTTP Request): Fetches the individual review page HTML.
• Extract review (HTML): Extracts the main article content of a single review.
• Information Extractor (Langchain): Uses DeepSeek to parse the HTML and extract structured metadata from the review text.
• Sentiment Analysis (Langchain): Uses OpenAI to analyze the sentiment of the extracted review text.
• Update sheet (Google Sheets): Appends or updates the processed review data, including extracted metadata and sentiment, to the Google Sheet.
• Sticky Notes: Provide guidance and hints for configuration.