Auto-Generate and Post LinkedIn Carousels with n8n, Gemini AI, and Post Nitro

Input: TechRadar RSS or manual form title/description → Process: Gemini AI filters news and generates carousel copy → Output: Post Nitro PDF + automated LinkedIn post

Who

LinkedIn operators or creators who need regular professional content and run a self-hosted n8n instance

You get

Automatically fetch tech news daily, have AI select and generate carousel content, then produce a PDF and post it to LinkedIn

You need

n8n · post-nitro · linkedin

Steps

1. Trigger the workflow

   The workflow has two start paths: 1. Automatic: Schedule Trigger node set to fire at 6:00 AM daily. 2. Manual: Form Trigger node provides a form with Title and Description fields. Activate the workflow in n8n for the automatic trigger to take effect; for manual triggering, visit the form path and submit correctly filled fields.

2. Fetch and parse news

   On the automatic path, the TechRadar News node sends an HTTP GET to https://www.techradar.com/feeds.xml to fetch the RSS feed. The Parse HTML node (Code node) runs JavaScript to parse the RSS XML: - Uses regex to extract all <item> tags - Parses title, description, link, pubDate, category, dc:creator, media:content fields - Strips HTML tags and decodes entities - Outputs a structured JSON array to result This code is tailored to the TechRadar RSS format; changing the source requires modifying the parsing logic.

3. AI selects relevant news

   The Decide which News to Choose node (Basic LLM Chain) connects to the Gemini Flash 2.5 model, feeds it the parsed news list, and has AI pick the single most relevant item. Node parameter: promptType=define, with the prompt text defined directly.

   Execution prompt · Source text。Decide which News to Choose node text parameter (Basic LLM Chain → promptType: define)

   Copy prompt

   =You are an AI assistant. Your only task is to select the single most relevant news item from the provided news title list.
   
   Relevance criteria (in order of importance):
   1. n8n Automation
   2. AI and Technology News
   3. Personal brand development and motivation
   
   Exclusion rules:
   - Ignore any news about love, relationships, entertainment gossip, or censored/inappropriate topics.
   
   News list (each item has a number and a title):
   ```
   [
   {{ JSON.stringify($json.result.map((item, index) => ({ number: index+1, title: item.title })), null, 2) }}
   ]
   ```
   
   Return format:
   - Return only the number of the most relevant news item.
   - Do not return text, explanations, or anything else. Just a number
   
   Example valid responses:
   ```
   6
   ```

4. Generate carousel content

   The Get Title and Description node (Code) merges title and description from either automatic or manual input, outputting combinedTitle and combinedDescription. The Gemini Generate node (Google Gemini) uses the gemini-2.5-flash model to generate LinkedIn carousel copy based on the news title and description. Requires strict JSON output with starting_slide, body_slide (1–3 slides), and ending_slide. The Parse LLM Response1 node (Code) cleans any code fences and parses the JSON, extracting the slides array for downstream use.

   Execution prompt · Source text。Gemini Generate node messages.values[0].content parameter

   Copy prompt

   =You are an AI Agent that generates LinkedIn carousel post content.
   
   IMPORTANT: Return only raw JSON as plain text. Do not call functions, do not use tool calls, do not add extra text.
   
   Your task: Write engaging, professional slide text for a LinkedIn carousel based on the provided title and description.
   
   News Input:
   ```
   Title: {{ $json.combinedTitle }}
   Description: {{ $json.combinedDescription }}
   ```
   
   Carousel Requirements:
   - Exactly 1 `starting_slide` → strong hook, with heading, sub_heading, description, and CTA button.
   - At least 1 `body_slide` → explain key insights or takeaways from the news (you may create 2–3 if relevant).
   - Exactly 1 `ending_slide` → motivational close with sub_heading, heading, description, and CTA button.
   - Keep tone professional, concise, and LinkedIn-appropriate.
   - Do not include irrelevant topics (love, politics, entertainment gossip).
   - Ensure headings are short and catchy; descriptions can be 1–3 sentences.
   - Output must be valid JSON only, matching the schema below.
   
   Response Format:
   ```
   {
   "slides": [
     {
       "type": "starting_slide",
       "sub_heading": "string",
       "heading": "string",
       "description": "string",
       "cta_button": "string"
     },
     {
       "type": "body_slide",
       "heading": "string",
       "description": "string"
     },
     {
       "type": "ending_slide",
       "sub_heading": "string",
       "heading": "string",
       "description": "string",
       "cta_button": "string"
     }
   ]
   }
   ```

5. Generate PDF and download

   The ImportSlides embedPost node (Post Nitro) receives the slides JSON and calls the importSlides operation to generate a carousel PDF. - Requires brandId and templateId (hardcoded example values in the template; replace with your own Post Nitro credentials) - Output includes a PDF download URL The Download PDF node (HTTP Request) fetches the PDF from the Post Nitro return URL, saving it as binary my.pdf.

6. Upload PDF to LinkedIn

   The Linkedin User URN node (HTTP Request) retrieves the current user's sub identifier via the LinkedIn API. The Get Upload URL node (HTTP Request) initializes an upload with LinkedIn's document endpoint to obtain an uploadUrl and document URN: - POST https://api.linkedin.com/rest/documents?action=initializeUpload - Header: LinkedIn-Version: 202307 - Body: owner uses urn:li:person:{{sub}} from the previous step The Upload PDF node (HTTP Request) uploads my.pdf via PUT to the obtained uploadUrl: - Header: Content-Type: application/pdf - Body: binary data On success, the document URN will be used in the final post.

7. Generate copy and publish

   The Gemini Generate1 node (Google Gemini) uses gemini-2.5-flash to generate LinkedIn post body copy from the original title and description. The Post to LinkedIn node (HTTP Request) publishes the final post: - POST https://api.linkedin.com/rest/posts - Headers: LinkedIn-Version: 202408, Content-Type: application/json - Body includes author, commentary (AI-generated copy), visibility=PUBLIC, content.media (title + uploaded document URN), lifecycleState=PUBLISHED After publishing, the post appears on the LinkedIn home feed.

   Execution prompt · Source text。Gemini Generate1 node messages.values[0].content parameter

   Copy prompt

   =You are an AI Agent that generates LinkedIn post description.
   
   IMPORTANT: Return only raw text.
   
   Your task: Write engaging, professional text for a LinkedIn post based on the provided title and description.
   
   News Input:
   ```
   Title: {{ $('Get Title and Description').item.json.combinedTitle }}
   Description: {{ $('Get Title and Description').item.json.combinedDescription }}
   ```
   
   Example Response Format:
   ```
   🚀 Just launched my first fully automated LinkedIn carousel — created end-to-end by my AI Employee (built in n8n).
   
   No manual editing, no wasted time. The system handles content creation → carousel design → posting all on its own.
   
   ✨ I’m offering to set this up for 5 people free of cost. If you want to automate your content creation and save hours every week, drop a comment below.
   ```

You are an n8n workflow setup assistant. The user wants to reproduce a template that automatically fetches tech news, generates LinkedIn carousel content, and publishes it. Help the user:
1. Confirm they have a self-hosted n8n instance (community nodes are incompatible with n8n Cloud)
2. Verify the @postnitro/n8n-nodes-postnitro-ai community node is installed
3. Prepare three API credentials: Google Gemini, Post Nitro (API key + Template ID + Brand ID), and LinkedIn OAuth2
4. After importing from the template marketplace, replace all hardcoded credential IDs, brandId, and templateId
5. When testing the form trigger, ensure Title and Description are filled correctly (no input validation exists)
6. Suggest running manually first to verify each node's output, then activate the scheduled trigger

Note: The brandId and templateId in the template are the author's example values and must be replaced with your own Post Nitro configuration.