Paperless-ngx: So sortieren sich deine Dokumente endlich von selbst (KI-Workflow)

Automatically tag documents in Paperless-ngx with n8n and local AI You can download the complete n8n flow for free from the linked blog post. It's self-hosted, without paperless-ai. In the video, I demonstrate how an n8n flow automatically tags new documents in Paperless-ngx. A locally running language model using Ollama assigns titles, tags, and the appropriate document type. You retain control and remain independent of external tools whose development might eventually stall. Your n8n flow is ready for you to use. Your n8n flow is ready for you to use automatically. The ... A locally running language model using Ollama assigns titles, tags, and the appropriate document type. You retain control and remain independent of external tools whose development could eventually stall. The n8n flow is ready to use for Paperless-ngx. Blog post with n8n-Flow for download (free): https://www.digitalisierung-mit-kopf.... AI course (host your own LLM server): https://www.digitalisierung-mit-kopf.... Chapters 00:00 Intro 00:00 Why not paperless-ai 00:00 What the Flow does 00:00 Importing n8n-Flow 00:00 Setting up the paperless webhook 00:00 Creating API keys and credentials 00:00 Ollama integration and model selection 00:00 System Prompt explained 00:00 Live test with a real-world example Document 00:00 Extension Ideas 00:00 Outro Tools Used paperless-ngx: https://docs.paperless-ngx.com n8n: https://n8n.io Ollama: https://ollama.com Model in the video: gpt-oss:20b The System Prompt from the Video As promised in the video, here is the complete system prompt for you to copy. Adapt the tags, document types, and sample data to your own situation. Role You are a document classifier for paperless-ngx. Analyze the full text of a document and derive the following values: Title (String) Tags (List of IDs as Int) Document Type (ID as Int) Title Assign a meaningful title that concisely summarizes the document. Example: IHK Membership Fee Notice coursemine GmbH Zalando Reminder Property Tax Assessment Sample Street 1 Munich Permitted Tags (multiple allowed, ID see data): Important Rule: Only add tags if they are relevant. Do not add tags if none are relevant. Company If the document refers to "coursemine GmbH". Only then! Not for other companies! Example: Incoming invoice with recipient coursemine GmbH Rental If the document relates to at least one of our rental properties or concerns our rental business in general. Properties are exclusively: Sample Street 1 Munich Example Road 3 Landsberg am Lech Health If the document has a medical relevance. Example: Lab results, doctor's letter, etc. Permitted Document Types (only one allowed, ID see data): Important Rule: Only add if applicable. No document type if none matches: Invoice Invoices, receipts, etc. Correspondence Letters. Not: Attachments that might have been part of a letter (because then everything would be considered correspondence). Utility Bill Utility bill, heating bill, service charge statement, etc. Output Format Output strictly JSON (without a JSON prefix) in the following keys: title, tags, doctype The rule "Only set tags or document types if they match" is important. Without this rule, local models tend to tag every document with everything in the prompt. Community and Support If you have questions about the flow or get stuck somewhere: Viewers help each other in the forum. https://forum.digitalisierung-mit-kop... More from Digitalization with Brains Subscribe to the channel:    / channel   Website: https://www.digitalisierung-mit-kopf.de Newsletter (10 Euro discount on courses): https://www.digitalisierung-mit-kopf.de paperless-ngx Online Demo: https://www.digitalisierung-mit-kopf.... About the channel "Digitalization with Brains" is all about the paperless office, especially with paperless-ngx and Obsidian. For individuals and freelancers. #paperlessngx #n8n #Ollama #paperlessoffice #selfhosting