Self-hosted webv0.21
saseo
Literature system that decomposes PDFs into markdown and images, stored and searched by library
764
Automated tests
9
Document types auto-detected
1,337
Records reclassified, 0 failures
Highlights
Originals stay untouched
PDFs are never modified; markdown, images and indexes can be rebuilt any time
Blank over guess
Weakly supported metadata is left empty and sent to a review queue
Readable without a server
Reading a library folder's manifest is enough to get all of its metadata
Pipeline
- 1PDF uploadDeduplicated
- 2ConversionDocling / OCR
- 3Metadata extractionWeb enrichment
- 4Commit gateGaps go to review
- 5Librarymanifest
- 6Search & read
Stack
FastAPIPostgreSQLReactTypeScriptDocker