n8n Google Drive → Pinecone Sync with OpenAI Embeddings

Keep your Pinecone knowledge base perfectly synced with a Google Drive folder—automatically

This n8n workflow extracts text from files in a specified Google Drive folder, generates OpenAI embeddings, and upserts them into Pinecone with helpful metadata. It also tracks what’s been indexed, removes vectors for deleted files, and emails an admin via Gmail when unsupported formats are encountered.

What this workflow does

• Triggers: Runs manually, on a daily schedule (3:00 AM), or when a new file is created in the configured Google Drive folder.
• Scans Drive: Lists files in the target folder, skips temporary conversion/OCR artifacts, and processes files one at a time.
• Index tracking: Uses an n8n Data Table to skip files already indexed and unchanged since the last recorded modified time.
• File content extraction based on MIME type:
  • Exports Google Docs/Sheets for text extraction
  • Extracts text from PDFs (including OCR-conversion for scanned PDFs)
  • Parses Excel files
  • Describes images using OpenAI Vision
  • Converts DOCX/PPTX to Google formats before export
• Embedding + Pinecone sync:
  • Splits extracted text into chunks
  • Creates OpenAI embeddings
  • Deletes existing Pinecone vectors for the same fileId
  • Upserts updated vectors with metadata (title, source, fileId)
• Cleanup: Updates the tracking table in the Data Table and removes Pinecone vectors + entries for files deleted from Drive.
• Unsupported types: Sends a Gmail notification to the admin and records the skip to prevent repeated alerts.

Use cases

• Automatically powering a RAG/semantic search knowledge base from company documents stored in Google Drive
• Keeping Pinecone embeddings current as documents are updated, added, or removed
• Monitoring and triaging unsupported file formats via Gmail alerts

Technical details

• Integrations/Nodes: Google Drive, Gmail, n8n Data Table, and logic nodes including if, set, code, and no op
• Core flow: MIME detection → content extraction (incl. OCR and Vision for images) → chunking → OpenAI embeddings → Pinecone upsert with metadata