This project provides an AI-powered enhancement for paperless-ngx, automating document titling, tagging, and metadata generation. It targets users of paperless-ngx seeking to improve document organization and text extraction accuracy, offering significant time savings and more intelligent document management.

How It Works

paperless-gpt leverages Large Language Models (LLMs) and LLM Vision capabilities to process documents. It can use OpenAI or Ollama for LLM-based OCR, offering superior accuracy on challenging scans compared to traditional OCR. Alternatively, it integrates with enterprise-grade OCR services like Google Document AI or Azure Document Intelligence. The tool generates context-aware titles, tags, and correspondent information, and can create searchable PDFs with accurate text layers.

Quick Start & Requirements

• Installation: Docker Compose is the recommended installation method.
• Prerequisites: Docker, a running instance of paperless-ngx, and access to an LLM provider (OpenAI API key or Ollama server).
• Configuration: Requires setting environment variables for paperless-ngx connection details and LLM provider configuration.
• Documentation: Detailed configuration options and usage examples are available in the README.

Highlighted Details

• LLM-Enhanced OCR: Utilizes OpenAI or Ollama for improved text extraction accuracy.
• Multiple OCR Providers: Supports LLM-based OCR, Azure Document Intelligence, Google Document AI, and Docling Server.
• Advanced PDF Features: Generates searchable PDFs with accurate text layers (currently only with Google Document AI).
• Automatic Metadata: Generates titles, tags, and correspondent information.
• Customizable Prompts: Allows users to tailor AI behavior via prompt templates.

Maintenance & Community

The project is actively maintained by icereed. Further community engagement details are not explicitly provided in the README.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Compatible with paperless-ngx. The MIT license permits commercial use and linking with closed-source applications.

Limitations & Caveats

Enhanced PDF features, including text layer generation and hOCR integration, are exclusively supported when using Google Document AI as the OCR provider. The PDF_REPLACE: "true" option for overwriting original documents is flagged as potentially dangerous and requires extreme caution due to the risk of data loss.