Paperless-AI: LLM Tagging and Document Chat for Paperless-ngx
Paperless-AI connects Paperless-ngx to a large language model. As documents arrive, it reads them and assigns titles, tags, document types, and correspondents, and it adds a chat interface for finding documents by asking questions in plain language.
Status: the maintainer describes the project as currently unmaintained while they consider a rewrite, noting that Paperless-ngx is working on official AI integration of its own. It still works for many users, but expect no fixes for now. paperless-gpt is the actively maintained alternative.
Features
- Automatic classification of new documents: title, tags, type, and correspondent
- RAG-based chat: ask “what was the warranty period on the dishwasher?” and get an answer with the source document
- Rules and filters to decide which documents are processed
- A manual mode for sensitive documents you do not want sent automatically
LLM providers
OpenAI, Ollama, DeepSeek, OpenRouter, Perplexity, Together.ai, Google Gemini, and OpenAI-compatible servers such as vLLM and LiteLLM.
Privacy
Documents in Paperless are often the most sensitive data you own: bank statements, medical letters, tax records. Sending them to a cloud LLM sends their contents to that provider. Ollama keeps everything on your hardware; our Ollama guide covers running models locally.
Deployment
It runs as a Docker container alongside Paperless-ngx, using a Paperless API token and your LLM provider settings.
License
MIT.