Local Document Intelligence Pipeline That Unifies OCR, AI Classification, and Search

desktop app real project •• multiple requests

Getting a complete document intelligence workflow running locally requires stitching together Paperless-ngx for storage, Stirling PDF for manipulation, paperless-gpt for AI tagging, and custom scripts for the gaps. Built-in OCR still fails on tables and photographs. Users want one self-hosted pipeline that handles scan-to-searchable-archive with AI categorization without uploading anything to the cloud.

builder note

Don't rebuild Paperless-ngx. Build the missing middle layer: a local OCR+AI service that accepts documents via API, runs vision-model OCR (not Tesseract), classifies, extracts structured data, and pushes results back to Paperless-ngx or any document store. Ship it as a single Docker container with Qwen-VL or similar baked in.

landscape (3 existing solutions)

The pieces exist but the pipeline is fragmented across 3-4 separate tools requiring Docker expertise to glue together. The approaching native AI in Paperless-ngx may close part of this gap, but the OCR quality problem (tables, photos, handwriting) persists because Tesseract is the bottleneck. Vision-capable local LLMs are the solution but integration is DIY.

Paperless-ngx Excellent document management but built-in Tesseract OCR fails on tables, photos, and complex layouts. AI integration is bolted on via third-party tools, not native. Official AI integration is coming but timeline unclear.

Stirling PDF PDF manipulation powerhouse with OCR support, but it's a tool, not a pipeline. No automatic classification, no persistent document store, no search index.

paperless-gpt / paperless-ai Bridges the AI gap for Paperless-ngx but requires separate deployment, configuration, and maintenance. PDF text layer generation only works with Google Cloud AI, defeating the local-only purpose.

sources (3)

other https://github.com/icereed/paperless-gpt "LLM Vision OCR to handle paperless-ngx documents" 2026-03-01

other https://github.com/paperless-ngx/paperless-ngx/discussions/5... "Alternative OCR engines requested for better accuracy" 2026-01-20

other http://www.blog.brightcoding.dev/2026/01/16/offline-ocr-revo... "offline OCR revolution transforming local document processing" 2026-01-16

self-hostedOCRAIdocumentsprivacy