🔬 Nested PDF analysis with Mistral AI & OneDrive for document summarization
⚡ 720 views · 🔬 Document Extraction & Analysis
Description
Description
Automate the discovery and analysis of PDF files across a deeply nested OneDrive folder structure. This workflow recursively searches folders, filters for new or updated PDFs, extracts text, and uses a Mistral AI agent to generate a concise Executive Summary, Key Findings, and Structured Metadata (Date, Location, etc.), storing all insights into a n8n Data Table for easy access and further automation.
Key Features & How It Works
Scheduled Trigger & Recursive Folder Search:
-
The workflow runs automatically (scheduled for 8 PM in this template) to monitor a specified main folder on OneDrive.
-
It performs a deep, multi-level search (up to 8 layers) across subfolders to ensure no documents are missed.
Smart Deduplication & Filtering:
-
It checks new files against an internal n8n Data Table using the Compare Datasets node, ensuring only new or unique PDF files are processed, saving AI credits and processing time.
-
A size check is also included, preventing attempts to process excessively large files.
-
AI-Powered Document Intelligence (Mistral LLM):
For each new PDF, the workflow extracts the text and passes it to a Mistral AI model for dual-stream analysis:
-
Overview Agent: Generates an impartial, professional Executive Summary, a list of Key Findings & Data Points, and the document’s Scope/Context.
-
Document Information Agent: Extracts crucial metadata, including the single most relevant date, location (City/State/Country), and professional information (Name, Title, Organization).
Structured Output and Archiving:
-
AI outputs are meticulously validated and reformatted into a clean JSON object using Structured Output Parsers.
-
The complete analysis, along with the original file name and path, is then logged as a new row in an n8n Data Table.
Setup Notes
OneDrive Folder: You must specify the exact name of your main folder in the ‘Search for Main Folder’ node.
Data Table: Ensure your n8n Data Table exists with the required columns: Summary, Key_Findings, Scope, Date, Location, File_Name, and Path.
Deep Folder Structure: The current configuration supports up to 8 levels of subfolders. If your files go deeper, you may need to add more “Get items in a folder” and “If” nodes.
AI Customization: Review the AI agent prompts and the structured output schemas to customize the fields you want to extract or the summary style you require.
Extend This Workflow
The final output is organized data. You can easily extend this workflow to:
-
Send daily/weekly digest emails with new summaries.
-
Sync the extracted data to a Google Sheet, Airtable, or other database.
-
Add a secondary AI agent to perform follow-up actions based on the “Key Findings.”
🔗 Nodes Used
Microsoft OneDrive, Schedule Trigger, AI Agent, Basic LLM Chain, Structured Output Parser, Extract from File
📥 Import
Download workflow.json and import into n8n:
Workflow menu → Import from File