paperless-gpt
Use LLMs and LLM Vision (OCR) to handle paperless-ngx - Document Digitalization powered by AI
What is paperless-gpt?
Use LLMs and LLM Vision (OCR) to handle paperless-ngx - Document Digitalization powered by AI
Overview of paperless-gpt
What it is
paperless-gpt is an open-source tool that integrates with paperless-ngx to use Large Language Models (LLMs) and LLM Vision (OCR) for document digitalization. It generates AI-powered document titles, tags, and links related documents together by reading reference numbers such as invoice, contract, or case numbers. The project is maintained by Icereed and is available under the MIT license.
Key features
- LLM-Enhanced OCR: Uses OpenAI or Ollama for context-aware text extraction from scans.
- Links related documents: Automatically fills paperless-ngx Document Link fields by matching reference numbers (e.g., invoice, contract, case number) across documents.
- Supports specialized AI OCR services: Google Document AI, Azure Document Intelligence, and Docling Server.
- Automatic title, tag, correspondent, and custom field generation with configurable write modes (Append, Update, Replace).
- Searchable and selectable PDF generation with transparent text layers; can be uploaded as a new document version in paperless-ngx.
- Customizable prompts and AI workflows via a web UI, allowing per-document-type trigger tags, prompts, and processing steps.
- Ad-hoc document analysis using custom prompts on selected documents.
Use cases
paperless-gpt is designed for users who need to organize large volumes of scanned documents in paperless-ngx. It is suitable for automating the naming, tagging, and linking of invoices, contracts, reminders, amendments, delivery notes, and correspondence. It can also be used to extract text from low-quality or tricky scans using LLM-based OCR, and to generate searchable PDFs from image-only documents.
Getting started
See the repository README for detailed installation instructions. The project supports Docker Compose deployment with environment variables. A basic Docker Compose example is provided in the repository.
License and pricing
The project is licensed under the MIT license. It is free and open source. Managed hosting is available for Germany, Austria, and Switzerland through server.camp, which shares revenue with the project.
Based on the project's public README and repository metadata.