Hello Developers,
I’m working on a system that uses ChatGPT models for document redaction and report generation and would like guidance on best practices and architecture models and apis subscription details
Use Case – Document Redaction:
-
Input documents include PDFs (text-based and scanned) and image files.
-
Text is extracted using PDF parsing and OCR.
-
ChatGPT models are used to identify sensitive information such as emails, phone numbers, SSN/national IDs, credit card numbers, names, addresses, and organization-specific identifiers.
-
Redaction is config-driven, allowing entity types to be enabled or disabled per organization.
-
Identified entities are mapped back to document coordinates and visually redacted in the final PDF/image output.
Use Case – Report Generation:
-
Generate reports from redaction results using chat-based instructions.
-
Supported output formats: PDF, Excel, CSV, and JSON.
-
Reports include summaries, entity counts, and compliance-ready tables.
I’m looking for recommendations on:
-
Prompt design for consistent entity detection
-
Handling OCR inaccuracies
-
Best practices for redaction accuracy and auditability
-
Efficient generation of multi-format reports
Any insights, sample architectures, or experiences would be greatly appreciated.
suggest models need to use and api plan need to buy and also documentation forum link for the same.
Thanks in advance!