Templates / AI-powered document processing & organization with Gemini, VLM Run & Google Sheets

AI-powered document processing & organization with Gemini, VLM Run & Google Sheets

Unknown

by Atik ✓ · 557 views · created 2025-09-20 · Document Extraction, AI RAG

Unknown. This template uses community nodes we can't verify statically. Uses community nodes that can't be verified statically (unverified community packages are off by default in 3.0). Static check, last run 2026-10-19.

Community nodes

These come from third-party npm packages. In n8n 3.0 unverified community packages are off by default (N8N_UNVERIFIED_PACKAGES_ENABLED=false).

Nodes used (9)

Google DriveOpenAI Chat ModelHTTP RequestGoogle SheetsGoogle GeminiAI Agent

What it does

Automate multi-document handling with AI-powered extraction that adapts to any format and organizes it instantly. What this workflow does 1. Monitors Google Drive for new uploads (receipts, resumes, claims, physician orders, blueprints, or any doc type) 2. Automatically downloads and prepares files for analysis 3. Identifies the document type using Google Gemini 4. Parses structured data via the trusted VLM Run node with OCR + layout parsing 5. Stores records in Google Sheets — AI Agent maps values to the correct sheet dynamically Setup Prerequisites: Google Drive & Google Sheets accounts, VLM Run API credentials, n8n instance. Install the verified VLM Run node by searching for VLM Run in the node list, then click Install. Once installed, you can integrate it directly for high-accuracy data extraction. Quick Setup: 1. Configure Google Drive OAuth2 and select a folder for uploads 2. Add VLM Run API credentials 3. Create a Master Reference Google Sheet with the following structure: | DocumentName | SpreadsheetID | | ---------------------- | ----------------------------- | | Receipt | your-receipt-sheet-id | | Resume | your-resume-sheet-id | | Physician Order | your-physician-order-sheet-id | | Claims Processing | your-claims-sheet-id | | Construction Blueprint | your-blueprint-sheet-id | The first column holds the document type, and the second column holds the target sheet ID where extracted data should be appended. 4. In the AI Agent node, edit the agent prompt to: Analyze the JSON payload from VLM Run Look up the document type in the Master Reference Sheet If a matching sheet exists → fetch headers, then append data accordingly If headers don’t exist → create them from JSON keys, then insert values If no sheet exists → add the new type to the Master Reference with an empty Spreadsheet ID 5. Test with a sample upload and activate the workflow How to customize this workflow to your needs Extend functionality by: Adjusting the AI Agent prompt to support any new document schema (just update field mappings) Adding support for multi-language OCR or complex layouts in VLM Run Linking Sheets data to BI dashboards or reporting tools Triggering notifications when new entries are stored This workflow leverages the VLM Run node for flexible, precision extraction and the AI Agent for intelligent mapping, creating a powerful system that adapts to any document type with minimal setup changes.

View the original on n8n.io → Check your own copy

Get the weekly “what broke” email

New templates, newly broken ones, and n8n release changes. One short email a week.

Form not loading? Subscribe here.