Weekly screenshot check of web pages with a Linear report

Weekly screenshot check of web pages with a Linear report

An n8n workflow in two parts. Part A takes a base screenshot of every page on a Google Sheet. Part B runs each Monday, takes a fresh screenshot, lets Google Gemini compare the two and files the differences as one Linear issue.

What is inside

• workflow.json: the workflow as n8n exports it, thirty-four nodes. Twenty-four do the work and ten are sticky notes that explain the steps. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The chat model named models/gemini-1.5-pro-latest. Google's models and deprecations pages, read on 12 November 2026, no longer list any 1.5 model, so Mindket's copy uses models/gemini-3.6-flash, which the models page lists as available with image input.
• Get Webpages List still had a leftover filter that kept only row 2 of the sheet. Mindket's copy removes it, so Part B checks every row.
• The two screenshot nodes sent the Apify token as a token query parameter. Apify's API docs, read on 12 November 2026, call that less secure than the Authorization header, so Mindket's copy uses a Header Auth credential.
• The Drive folder, the Google Sheet and the Linear team belonged to the original author. Those fields are empty now, so pick your own.
• The sticky notes held referral links, a link to an example sheet and an invitation to a chat server. They are replaced by a plain link and a line about the sheet's columns.
• The file had no settings block, so Mindket's copy adds executionOrder v1, the setting n8n gives a new workflow.
• Several nodes were linked to credentials in the author's n8n. The links are gone, so select your own.
• Nothing else. The prompt, the loops and the report text are the file's own.

Before you start

• An n8n instance. Publish the workflow if you want the Monday run to happen by itself.
• A Google Sheets file with the columns service, url and base. Put a name in service and a full page address in url, one row for each page. Leave base empty, because Part A fills it.
• A Google Drive folder for the base screenshots.
• An Apify account and its API token. The screenshots come from Apify's apify/screenshot-url actor.
• A Google Gemini API key from Google AI Studio.
• A Linear account with a team to file the issue in.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Google Sheets credential. Select it in Get URLs with Missing Base Images, Update Base Image and Get Webpages List, and pick your file and sheet in each.
3. Add a Google Drive credential. Select it in Upload to Drive and Base Image, and pick your folder in Upload to Drive.
4. Add a Header Auth credential for Apify. Name it Authorization and give it the value Bearer <YOUR_APIFY_TOKEN> with your own token. Select it in Run Webpage Screenshot and Run Webpage Screenshot1.
5. Add a Google Gemini credential with your key and select it in Google Gemini Chat Model. Check that the model is models/gemini-3.6-flash.
6. Add a Linear credential, select it in Create Report and choose your team there.
7. Press Execute workflow. This runs Part A. Open the sheet afterwards and check that each row has a file id in base.
8. To try Part B at once, open Schedule Trigger and run that step. Publish the workflow to get the run every Monday at 6.

How a run goes

1. Part A starts from the manual trigger. Get URLs with Missing Base Images reads the rows of the sheet.
2. Loop Over Items sends each page to Run Webpage Screenshot, which runs the Apify actor and waits for its answer. The window is 1280 pixels wide, and Wait pauses two seconds between pages.
3. When the loop ends, Merge pairs each row with its screenshot. Download Screenshot fetches the picture, Upload to Drive saves it under the encoded page address and Update Base Image writes the Drive file id into base.
4. Part B starts from Schedule Trigger, which fires on Mondays at 6. Get Webpages List reads the rows, and For Each Webpage... takes them one at a time.
5. For each page, Base Image downloads the saved screenshot from Drive. Run Webpage Screenshot1 takes a new one, Download New Screenshot fetches it and Combine Screenshots joins the two.
6. Visual Regression Agent (Basic LLM Chain 1.4) sends both pictures to Google Gemini Chat Model as Image (Binary) messages. Its system message asks for changes in text, images, colours, position and layout. It says to ignore text styling and casing. Structured Output Parser turns the answer into a list of changes, each with a type, a description, a previous state and a current state. Wait1 pauses one second.
7. After the last page, Combine Row and Result joins each row with its list. Has Changes keeps only the pages with at least one change, and Aggregate gathers them. Create Report files a Linear issue named Visual Regression Report with the date. The issue lists each page and its changes.

Change the result

• Which pages: the rows of the sheet.
• When it runs: Schedule Trigger.
• What counts as a change: the system message in Visual Regression Agent.
• Where the report goes: Create Report. Swap it for a Slack or Jira node if you do not use Linear.
• Screenshot size and wait: the body of both screenshot nodes. Keep them the same, so the base and the new picture match.
• Model: modelName in Google Gemini Chat Model.

Good to know

• Despite its name, Get URLs with Missing Base Images reads every row. Each run of Part A saves a new base file for every page and rewrites base. The old files stay in Drive.
• Two screenshots of a live page rarely match exactly. Ads, carousels and dates can show up as changes, so read the first reports with that in mind.
• Apify's synchronous run endpoint ends a call after 300 seconds (Apify's API docs, read on 12 November 2026).
• Each page costs one Apify run and one Gemini call in Part B. Apify and Google bill them to your accounts.
• The page addresses go to Apify. Both pictures of each page go to Google, and the report text goes to Linear.
• The file holds no credential and no key.