4
[Feature Request] Self-hosted MinerU as a remote OCR engine
Source: paperless-ngx/paperless-ngx#12947 · opened by @Pacman1984
What MinerU is: [MinerU]( is an open-source, self-hostable document-parsing engine with a simple HTTP API (POST /file_parse). It turns PDFs and images into high-quality text/Markdown and handles complex layouts and tables far better than Tesseract. Why this instead of the current approach: paperless-ngx already has a "remote OCR" engine, but the only option is Azure Document Intelligence - a paid cloud service. MinerU is the same idea (offload OCR to a stronger engine) while staying 100% local and free, which fits the self-hosted spirit of paperless and avoids sending documents to the cloud. What I have: a PR that adds mineru as a remote OCR engine, deliberately built to align as closely as possible with the existing Azure implementation - same RemoteDocumentParser and PAPERLESS_REMOTE_OCR_* settings, no new dependencies, with backend tests and docs. Enable via PAPERLESS_REMOTE_OCR_ENGINE=mineru and PAPERLESS_REMOTE_OCR_ENDPOINT=<MinerU API URL> (no API key). …
No pledges yet. Be the first to back this.
Comments
Similar requests
[Feature Request] Make Remote OCR respect `PAPERLESS_OCR_MODE` and local preprocessing options
5 votes · 0 comments
[Feature Request] Support a self-hostable remote OCR service
30 votes · 0 comments
[Feature] Optional removal of blank pages from the archive file (originals untouched)
1 vote · 0 comments
[Feature Request] OCR: Extract specific data from zones into Fields [PR incoming]
12 votes · 0 comments
[Feature Request] OCR rendition on linked documents
2 votes · 0 comments
No comments yet.