Automated Document Verification via n8n
Budget: €18 – €36 EUR
Create an Automated Document Verification Workflow Using n8n
Objective:
Develop an automated workflow on the n8n platform to handle PDF or image documents submitted by users, perform OCR text extraction, verify extracted content using multiple AI language models through a voting system, and produce a clear verification result (valid or invalid). Additionally, specific fields must be extracted from the uploaded documents.
Workflow Steps:
File Submission:
Accept document uploads (formats: PDF, JPG, PNG, JPEG) via an HTTP webhook or direct upload.
OCR (Text Extraction):
Automatically extract text from uploaded documents.
Recommended services: Google Cloud Vision, AWS Textract, Tesseract OCR via API (select based on ease and accuracy).
Field Extraction:
Extract specific, predefined fields from the OCR-extracted text (e.g., name, date, amount, tax ID).
Text Verification Using AI Models (Voting System):
Send extracted text to three separate, cost-effective AI language models (e.g., GPT-3.5, smaller HuggingFace models, or other economical models).
Implement a voting mechanism: If all three cheaper models agree, produce a final result based on their consensus.
If discrepancies or irregularities arise (no unanimous decision), escalate verification to a more powerful and expensive model (e.g., GPT-4) for a final decision.
Result Handling:
Clearly indicate the verification result (e.g., "Document is valid" or "Document requires review").
Optionally, store extracted fields and verification results in a database or send them via email to a specified recipient.
Error Handling and Notifications:
Implement error handling for file upload issues, OCR extraction failures, and AI API errors.
Notify a specified administrator via email or Slack if an error occurs.
Deliverables:
Fully configured and operational n8n workflow.
Clear documentation explaining each step of the workflow and how to adjust API keys or endpoint URLs.
Video or written instructions demonstrating workflow operation.
Objective:
Develop an automated workflow on the n8n platform to handle PDF or image documents submitted by users, perform OCR text extraction, verify extracted content using multiple AI language models through a voting system, and produce a clear verification result (valid or invalid). Additionally, specific fields must be extracted from the uploaded documents.
Workflow Steps:
File Submission:
Accept document uploads (formats: PDF, JPG, PNG, JPEG) via an HTTP webhook or direct upload.
OCR (Text Extraction):
Automatically extract text from uploaded documents.
Recommended services: Google Cloud Vision, AWS Textract, Tesseract OCR via API (select based on ease and accuracy).
Field Extraction:
Extract specific, predefined fields from the OCR-extracted text (e.g., name, date, amount, tax ID).
Text Verification Using AI Models (Voting System):
Send extracted text to three separate, cost-effective AI language models (e.g., GPT-3.5, smaller HuggingFace models, or other economical models).
Implement a voting mechanism: If all three cheaper models agree, produce a final result based on their consensus.
If discrepancies or irregularities arise (no unanimous decision), escalate verification to a more powerful and expensive model (e.g., GPT-4) for a final decision.
Result Handling:
Clearly indicate the verification result (e.g., "Document is valid" or "Document requires review").
Optionally, store extracted fields and verification results in a database or send them via email to a specified recipient.
Error Handling and Notifications:
Implement error handling for file upload issues, OCR extraction failures, and AI API errors.
Notify a specified administrator via email or Slack if an error occurs.
Deliverables:
Fully configured and operational n8n workflow.
Clear documentation explaining each step of the workflow and how to adjust API keys or endpoint URLs.
Video or written instructions demonstrating workflow operation.