Ingest LinkedIn Posts from Google Drive to Pinecone (Gemini, Supabase)

Turn LinkedIn post screenshots & PDFs into searchable writing patterns—automatically

This n8n workflow watches a Google Drive folder for new LinkedIn post files, extracts the text (including OCR for images), uses Google Gemini + Groq to identify reusable writing patterns, embeds them with Gemini, and upserts the results into a Pinecone index—while logging processed files to Supabase to prevent duplicates.

What this workflow does

• Monitors Google Drive every minute for newly created files in a selected folder (e.g., /reference-content).
• Deduplicates using Supabase by checking whether the Google Drive file ID was already processed; if it exists, the workflow stops.
• Downloads the file and routes by MIME type:
  • PDFs: extracts text directly.
  • Images (JPEG): runs Tesseract OCR, then uses Groq to clean and reconstruct the OCR text.
• Extracts structured “writing patterns” with Google Gemini by converting normalized post text into JSON fields such as hook type, structure, tone, formatting notes, CTA type, and anti-patterns.
• Creates chunked documents, generates Gemini embeddings, and upserts them into a Pinecone vector index with metadata.
• Logs ingestion to Supabase (source file ID, file name, timestamp) so future runs skip the same file.

Use cases

• Build a Gemini-driven “LinkedIn swipe file” by continuously ingesting posts stored in Google Drive.
• Create a Pinecone-backed semantic search experience for marketers and SaaS teams.
• Generate structured playbooks (hooks, tone, CTAs, formatting notes) from saved LinkedIn content for faster writing.

Technical details

• Integrations: Google Drive, Supabase, Groq, Google Gemini (chat + embeddings), Pinecone.
• n8n logic & nodes: set, no op, merge, filter, switch, Supabase steps for duplicate checks and ingestion logging.
• OCR: Tesseract for images; PDF text extraction for documents.