Doctra — Document Parser
Parse PDFs, extract tables/charts, preview markdown, and download outputs.
VLM Provider
100 400
0 1
0 13
0 3
Target
VLM Provider
100 400
0 1
Restoration Task
Device
100 400
📄 Original PDF
✨ Enhanced PDF
Restoration Task
Restoration Device
VLM Provider
100 400
100 400
0 1
0 13
0 3
📄 Original PDF
✨ Enhanced PDF
Tips
- On Spaces, set a secret
VLM_API_KEYto enable VLM features. - Use Enhanced Parser for documents that need image restoration before parsing (scanned docs, low-quality PDFs).
- Use DocRes Image Restoration for standalone image enhancement without parsing.
- DocRes tasks:
appearance(default),dewarping,deshadowing,deblurring,binarization,end2end. - Outputs are saved under
outputs/<pdf_stem>/. - Note: Google Gemini VLM may not be available due to dependency conflicts. Use OpenAI, Anthropic, or other VLM providers.