Generate LinkedIn posts from books using OpenAI, LangChain & Pinecone vector search

Automates the process of generating, storing, and publishing engaging LinkedIn posts derived from books (PDFs) using AI and vector search.

🧠 Overview

This workflow:

1. Watches a Google Drive folder for new or updated book PDFs.
2. Extracts and embeds the content using OpenAI.
3. Stores the data in a Pinecone vector database.
4. Uses a LangChain agent to generate post ideas.
5. Creates concise LinkedIn posts with hook, insight, CTA.
6. Updates a Google Sheet and posts to LinkedIn.

🛠 Workflow Breakdown

📥 1. Google Drive Trigger

• Trigger: Watches a folder for new or updated PDF files.
• Action: Downloads the updated PDF.

📄 2. Extract and Embed Content

• Extract from File: Parses PDF to extract text.
• Text Splitter: Breaks text into chunks.
• Embeddings (OpenAI): Converts chunks into vector embeddings.
• Pinecone Vector Store: Saves the embeddings with the book name as namespace.

🧠 3. Post Idea Generation (LangChain Agent)

• Uses a prompt to:
  • Search Pinecone DB
  • Extract insights
  • Format into 5 LinkedIn post ideas with:
    • Hook
    • Insight
    • CTA
• Memory buffer and structured output parser are used for clean AI interaction.

✍️ 4. Post Creation

• Each idea is:
  • Split
  • Rewritten with a GPT model prompt to match LinkedIn tone
  • Styled for under 600 characters
  • Includes emojis, hashtags, and tone guidelines

📊 5. Google Sheet Integration

• Saves all generated posts to a Google Sheet.
• Marks status: "published" or "no".

🔁 6. Scheduled Publishing

• Every day:
  • Pulls an unpublished post
  • Publishes it to LinkedIn
  • Updates the post's status and timestamp in the Google Sheet

⚙️ Setup Guide

📂 Google Drive

• Create a folder for book PDFs
• Connect your Google Drive account to n8n
• Provide access token with file read permission

📊 Google Sheets

• Create a Google Sheet with columns:
  • bookname, hook, insight, cta, postContent, published, date
• Add credentials in n8n with read/write permission

🧠 Pinecone

• Set up a Pinecone project and index (linkdenpost)
• Namespace will be auto-named using the book filename

🔑 API Credentials Required

• OpenAI API (for embeddings and post generation)
• Pinecone API (for vector storage and retrieval)
• LinkedIn OAuth2 (to publish posts)
• Google Drive & Sheets credentials

🔁 Flow Summary

graph TD A[Google Drive Trigger] --> B[Download PDF] B --> C[Extract Text] C --> D[Text Splitter] D --> E[Create Embeddings] E --> F[Pinecone Vector Store] F --> G[LangChain Agent] G --> H[Structured Output (5 Post Ideas)] H --> I[Split Ideas] I --> J[Format as LinkedIn Post (GPT)] J --> K[Store in Google Sheet] L[Schedule Trigger] --> M[Get Unpublished Post] M --> N[Post to LinkedIn] N --> O[Mark as Published]

🧪 Prompt Example (Used in LangChain Agent)

You are a content strategist. Search the Pinecone vector DB containing a book. Generate 5 unique LinkedIn post ideas with: - A Hook (curiosity driven) - Insight (summary < 100 words) - CTA ("Agree or disagree?", etc.) Respond in structured JSON: [ { "Hook": "...", "Insight": "...", "CTA": "..." }, ... ]

✅ Output Sample

{ "Hook": "Why your lab's results might be invalid 😱", "Insight": "ISO/IEC 17025 stresses that labs must plan and address risks to impartiality and validity.", "CTA": "Does your lab audit for these risks?" }

📆 Schedule Control

• Uses Schedule Trigger to post daily at a set time.
• Ensures automation with LinkedIn and accurate Google Sheet syncing.

📝 Notes

• Posts remain professional and concise for a LinkedIn audience
• Works with any PDF book
• Supports multi-book pipelines
• You can filter and tag books by filename or folder for segmenting post styles