scanbutler ghcr.io/tom-joad/scanbutler:latest https://ghcr.io bridge sh false https://github.com/Tom-Joad/scanbutler https://github.com/Tom-Joad/scanbutler https://raw.githubusercontent.com/Tom-Joad/unraid-templates/main/templates/scanbutler.xml https://raw.githubusercontent.com/Tom-Joad/unraid-templates/main/icons/scanbutler.png Turns scanned paper into searchable PDFs, one per document, named by topic and date. The Stacks input splits large scans (hundreds of pages, no separator sheets) into documents by content. The Scanner input takes files from a document scanner, one document per file, and gives them the same OCR and naming. Uses Mistral OCR (batch API) and a Mistral chat model; the text layer comes from a fresh Tesseract pass. No web UI: progress is in the container log, and a review.md per stack in the Work folder lists uncertain splits. An optional third input only adds the text layer and uploads to Paperless-ngx. Optionally reports the queue to a webhook such as Home Assistant. Needs a Mistral API key. Productivity: Tools: --security-opt no-new-privileges --memory=4g